Home

PDF file

1. essssosssosssosssonsonsnnssnnssnnsnnnsnnnsnnnsnnnsnnnsnnnsnnnnne 24 Maultiobiective OD MEN 22H on ee 25 Introduction and Definit 008 02 aA 25 Solutions obtained with the use of weighted sums in optimization procedures 21 Generation of uniform and random importance factors for multiobjective ODE ZA ON ee es eer ere aera 28 Decision MAKING tOO S senne ann A n R RE R 29 Template based Inverse Plannin ssesssesesessesssessesssessesseosseossesseesseesseosees 31 Post Implant ObHM1220 00 i neat 33 The User IntetfaCe eeeseeeseeseeseesoesseeseeseesoesseseeeseesoesoeseeesoesoesoeseeeseeseesosseeesesseeso 34 The FILE T O Daloes n ae Re 36 The Dose Parameter Dial otc vccsssstecestecaatiestete ie ETN 37 The Template based Inverse Planning Dialog e seessseesossesessossesessossesessesesses 40 Template VIEN O esos 48 Phe Optimization Dialea aussen naar 52 Post Plan OBtimization une ernennen 53 Template based inverse Plan Optimization enenenene 71 The Dose distribution Dial68 uuanaa ken a aa 72 Examoles of using WinOpt HDR sesssesessnensnsnssseenenenenensenenenenenenenennen 80 PTV based Sptimization meth d nu rin 80 DVH based multiobjective optimization method ssssssssssssnnssssnssssnnnssnnnnssnnnnnnsnnnnnne 83 Example of template based manual Inverse Plannins eeeee 86 WinOdt H DR Errors and Warninss eessssesesessnsnsesesenensnnenenene
2. 5 ave List Save the current list in a file Show constraints setting dialog Display dialog with histograms Be Display dialog with Pareto set bi objective plots WinOpt HDR User Manual 68 Histograms 4 This dialog shows two histograms cumulative or differential for a selected VOI The histograms of marked solutions are shown with a blue color while all other histograms are in red Only solutions that satisfy the criteria set in the Constraints dialog are shown Display all histograms selected and not selected else only selected Display limits critical dose values and reference dose value Saves corresponding image as a TIFF file mA H Select histogram to be shown Diff DWH of PTY Cum DYH of PTY Diff DWH of SBody Cum DWH of 5Body Diff DYH urethra Cum DYH urethra N Set color for not selected solutions 5 Set color for selected solution WinOpt HDR User Manual 69 i I u m Lu Se _onstraints This dialog is used to set constraints which must be satisfied for all non dominated solutions of a multiobjective optimization run Set constraints set by the DVH and objectives values from the set of all solutions No Constraints Removes all constraints applied WinOpt HDR User Manual 70 iPareta Set CSOHigh k BT ee n En CSOHigh E PTV _ LOW This dialog is used to display two dimensional projections of the Pareto set In blue colo
3. Template y_ hickness WinOpt HDR User Manual 101 2 The VOI Charisma File Here is an example of the VOI file that contains information about the VOIS such as the PTV and OARs The part that is not necessary for the optimization is marked The Charisma File formats have been specified and been developed by Prof Dimos Baltas The VOI Type must be one of the following types 1 CS for a organ at risk critical structure 2 BODY for a body which is assumed to include the PTV and the critical structures 3 PTV for the Target Contours of VOIs are assumed to be simple polygons with none of the points to be identical The first and last point of each contour are assumed to be connected i e the polygons are closed Not simple polygons i e without any intersection between any of its edges are not allowed Points with a difference of less that 1 100 mm in each coordinate can not be distinguished WinOpt HDR User Manual 102 C mia aia sla mia ul a si mia mia vll ac vl ul al u a u a wa al u al u al a vl a vl m a a a a vl a a a al a al S VOI Contours Data File CHARISMA R Vs 1 0 0 1 C opyrights Pi Medical Ltd All Rights protected PEER EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE EEE EEE EEE EEE HHT The Inner Surfaces or Contours of a Hole Type VOI are referred explicitly as INNER If no explicit reference to Inner Surface or Contours is made the OUTER surface and contours are meant PEER tt Hr tr rt HH HH HH HH HH HH HH HH HH HH
4. Moves the plane along the normal to the template This plane is shown in the anatomy window H Select the catheter density Use every fourth second or every possible catheter that can be used WinOpt HDR User Manual 51 Windpt HOR The position of the plane defined by the z slider is shown in the anatomy window WinOpt HDR User Manual 52 The Optimization Dialog This dialog is used to select the optimization method a template based inverse plan optimization or a post plan optimization It is used to set the parameters of the optimization methods start the optimization and finally to display the results If a multiobjective optimization is used a single solution can be selected using the decision making dialog Part of optimization dialog for template based post plan optimization WinOpt HDR User Manual 53 Post Plan Optimization Fast Simulated Annealing Apply a fast simulated annealing single or multiobjective dose optimization Apply a deterministic single or multiobjective dose optimization Multiabjective Evolutionary Apply an evolutionary single or multiobjective dose optimization Set Optimization Options an oe Set optimization options for selected optimization algorithm Start the optimization Interrupt current loop in optimization Show after optimization results and decision making tools for the selection of a single solution and for the analysis of the tra
5. WinOpt HDR User Manual 20 The gradient based deterministic algorithm can be extended to include dose variances for OARS In this case it is possible that the algorithm is trapped by local minima Joar UN or OD Doar XD Doar Where x is the theta step function No p is the number of sampling points in the i OAR and D 4r is the critical dose for the i OAR i e only dose values above the critical dose value for each OAR are considered in the objective function A fast simulated annealing algorithm is less sensitive to local minima and therefore is implemented in WinOpt HDR WinOpt HDR User Manual 21 Dose optimization using dose volume histogram based objectives In the past optimization methods were based on the dose of some sampling points at some special positions such as on the PTV contours or basal dose points The normalization of the dose was based on the values of these points For the dose points on the PTV contours for example the mean dose value was normalized to the prescription dose An optimization goal is the minimization of the variance For a solution with a very small dose variance the isodose of the prescription dose then coincides with a high degree with the PTV surface This normalization although imposes a constraint on the dose distributions and therefore to the obtained solution s If the PTV surface is complex it is impossible that the entire PTV receives a dose equal or larger than the prescription dose
6. 0597238 0 596103 0 595885 0 595748 0 594077 0 593349 0 589725 0 588031 MEAR dA 86 885000 86 600000 36 595000 36 335000 a6 245000 56 130000 55 405000 05 3595000 85 180000 85 100000 91 535000 82 025000 82 010000 84 800000 53 016000 61 265000 31 585000 80 300000 Ad ASENN 47 870000 48 365000 45 355000 48 605000 48 840000 48 890000 48 415000 48 345000 48 225000 43 135000 44 000000 44 675000 44 660000 47 905000 46 125000 43 540000 44 070000 42 555000 47 inn 0 075000 0 080000 0 080000 0 090000 0 125000 0 730000 0 295000 0 305000 0 325000 0 340000 4 350000 3 595000 3 620000 0 420000 1 370000 4 965000 4 285000 3 620000 NAAR 1 120000 0 395000 0 995000 0 655000 0 825000 0 805000 0 565000 0 555000 0 525000 0 515000 0 165000 0 175000 0 175000 0 445000 0 235000 0 140000 0 165000 0 130000 g aannnn Abbrechen This dialog displays all non dominated solutions of an optimization run The list includes the values of all objectives used in the optimization and the values of the DVHs of all solutions The whole list can be sorted for each item Solutions can be selected out of this list and marked The cumulative and differential DVHs of all VOIS of all solutions can be shown Constraints can be applied to the values of the objectives and the DVHs to filter only solutions that satisfy these constraints
7. 2 200000 2 300000 2 400000 2 500000 2 600000 2 700000 99 925000 98 260000 94 355000 87 975000 79 545000 69 285000 58 460000 47 800000 37 525000 27 755000 21 120000 16 555000 13 480000 10 985000 9 325000 8 080000 6 940000 6 065000 5 280000 4 700000 4 190000 9 700000 0 240000 9 800000 9 900000 0 235000 0 230000 WinOpt HDR User Manual 125 Statistical Parameters Mean ADEV Average deviation SDEV Standard deviation VAR Variance MIN MAX values High Statistics of Dose Distribution for the PTV Mean ADEV SDEV VAR MIN MAX 1 542382 0 444308 1 370637 1 878645 0 622747 129 869449 High Statistics of Dose Distribution for the BODY Mean ADEV SDEV VAR MIN MAX 0 230071 0 197105 0 370684 0 137407 0 031468 15 116395 High Statistics of Dose Distribution for the Surrounding Tissue Mean ADEV SDEV VAR MIN MAX 0 447663 0 157729 0 221776 0 049185 0 140545 11 782692 High Statistics of Dose Distribution for the Rectum Mean ADEV SDEV VAR MIN MAX 0 157791 0 067478 0 086317 0 007451 0 037300 0 532966 High Statistics of Dose Distribution for the urethra Mean ADEV SDEV VAR MIN MAX 1 246166 0 198082 0 237210 0 056269 0 574726 1 860300 Here follows statistical information obtained by the sampling points used in the Optimization Statistics for PTV Surface of Sampling points 208 Mean 1 0000 Average Deviation 0 1442 Standard Deviation 0 1759 Variance 0 0309 Skewness 0 1970 Kurtosis 0 5117 Minimum Index 4 Minimum
8. M Lahanas T Kemmerer N Mihckovic D Baltas N Zamboglon Optimized bounding boxes for three dimensional treatment planning in brachytherapy Med Phys 27 2333 2342 2000 Abstract Preprint in pdf format T Kemmerer M Lahanas D Baltas and N Zamboglou DVH computation comparisons using conventional methods and optimized FFT algorithms for brachytherapy Med Phys 27 2343 2356 2000 Abstract Preprint in pdf format Baltas D Mihckove N Giannonh S Lahanas M Kolotas C Zamboglou N New Tools of Three Dimensional Imaging Based Brachytherapy in the Frontiers of Radiation Therapy and Oncology Series 59 70 2000 Abstract K Karouzakis M Lahanas N Milickovic D Baltas and N Zamboglon Stratified sampling submitted for publication in Med Phys 2000 Abstract Preprint in pdf format M Lahanas N Mihckovic D Baltas and N Zamboglou Application of Multiobjective Evolutionary Algorithms for Dose Optimization Problems in Brachytherapy in Proceedings of the first international conference EMO 2001 Zurich Switzerland edited by E Zitzler K Deb L Thiele C A Coello Coello D Corne Lecture Notes in Computer Science Vol 1993 Springer 574 587 2001 Abstract N Mibhekovic M Lahanas D Baltas and N Zamboglou Comparison of Evolutionary and Deterministic Multiobjective Algorithms for Dose Optimization in Brachytherapy in Proceedings of the first international conference EMO 2001 Zur
9. Nucletron Columbia 184 192 1990 5 S Giannouh D Baltas N Mihckovic M Lahanas C Kolotas N Zamboglon N Uzunogin Autoactivation of Source Dwell Positions for HDR Brachytherapy Treatment Planning Med Phys 27 2517 2520 2000 Abstract Preprint in pdf format 6 C Kolotas D Baltas N Zamboglou CT Based Interstitial HDR Brachytherapy Strahlenther Onkol 9 175 419 427 1999 7 K Karouzakis M Lahanas N Mihckovic D Baltas and N Zamboglou Stratified sampling submitted for publication in Med Phys 2000 Abstract Preprint in pdf format 8 T Kemmerer M Lahanas D Baltas and N Zamboglou DVH computation comparisons using conventional methods and optimized FFT algorithms for brachytherapy Med Phys 27 2343 2356 2000 Abstract Preprint in pdf format 9 M Lahanas D Baltas N Mihckovie S Giannonh and N Zamboglou Generation of uniformly distributed dose points for anatomy based three dimensional dose optimization in brachytherapy Med Phys 27 1034 1046 2000 Abstract Preprint in pdf format 10 M Lahanas T Kemmerer N Mihckovic D Baltas N Zamboglou Optimized bounding boxes for three dimensional treatment planning in brachytherapy Med Phys 27 2333 2342 2000 Abstract Preprint in pdf format 11 M Lahanas D Baltas and M Papagiannopoulou Calculation of Dose Distributions in Brachytherapy using Monte Carlo Generated Dosimetric Look Up Tables MITTUG R
10. body with dose values at least D Vp that is covered by PTV It is also a measure of how much normal tissue outside the PTV is covered by D COIN can be calculated from the cumulative DVHs of the PTV and the body DVH pry and DVA respectively V is the volume of the body If D is chosen to be the reference dose D then the ideal situation is 1 COIN PTV DVH py DJ 100 Voy DVH oq D 5 ody We describe the dependence of the conformal index COIN on the choice of the reference dose value as the COIN distribution One objective in dose optimization in brachytherapy is the avoidance of excessive radiation inside the PTV and normal tissue This objective can be formulated in terms of the COIN distribution The integral of the COIN distribution from some defined value for example 1 5D can be used rep WinOpt HDR User Manual 30 The conformitv index is extended in the presence of organs at tisk Vics is the volume of the i critical structure and V cs D gt D erit is the volume of that critical structure that receives a dose that exceeds the critical dose level D y The product runs over all critical structures Noag In case where a critical structure receives dose above the critical value defined for that structure the conformity index will be reduced by a fraction that is proportional to the volume that exceeds this limit According to this Eq 2 is now extended to the following N i i OAR V D gt D ai COIN
11. scale i EEN 100 Ar ei Color s t Ea Abbrechen Here are the two marked solutions Nr 17 and 31 5 000 4500 4 000 4 500 a000 2 500 2 000 1 500 1 000 1 750 WindptHoR Here we se the isodose distribution for solution 31 The isodose for the prescription dose and 1 5 times the prescription dose is shown The dose is distributed such that it avoids the region around the urethra The PTV is shown solid in red WinOpt HDR User Manual 94 WinOpt HDR Errors and Warnings Optimization does not run Set the source parameters source strength and prescription dose WinOpt HDR 2 00 VOI contours which are not simple polygons crossings are reported As a result the ends firs and last contour is left open no triangulation Ie no sampling points are generated at the ends and the surface of these contours is not considered in the calculation of areas pu m m a pr 7 WAlin pt HDR V2 00 Points in VOI contours that are less than 0 01 mm and less than the minimum distance allowed in the ROIS file is found As a result the ends firs and last contour is left open no triangulation Le no sampling points are generated at the ends and the surface of these contours is not considered in the calculation of areas WinOpt HDR User Manual 95 Collision or contact warning between VOIS In the file WinOpt_Log txt see the list of triangles that are WinOpt HDR V2 00 in contact or intersect Collision Detectio
12. 0 0 0 0 0 0 0 0 0 0 4 50 0 0 0 4 0 6 0 0 4 0 0 4 00 0 0 4 0 0 0 0 0 0 0 0 3 50 0 0 0 0 10 0 6 0 0 4 0 3 00 0 0 T 0 0 0 0 10 0 0 0 2 50 0 0 0 0 10 0 0 0 0 4 0 2 00 0 0 6 0 0 0 0 9 0 0 0 1 50 0 0 0 5 0 0 0 0 3 0 0 1 00 0 0 0 0 0 2 0 0 0 0 0 FREE LENGTH cm A a B b C c D d E e F 8 00 out out out out out out out out out out out 7 50 out out out out out out out out out out out 7 00 out out out out out out out out out out out 6 50 out out out out out out out out out out out 6 00 out out out out out out out out out out out 5 50 out out out out out out out out out out out 5 00 out out out out out out out out out out out 4 50 out out out 9 800 out 10 050 out out 10 050 out out 4 00 out out 9 925 out out out out out out out out 3 50 out out out out 8 800 out 11 175 out out 9 825 out 3 00 out out 9 425 out out out out 9 050 out out out 2 50 out out out out 8 925 out out out out 10 050 out 2 00 out out 10 050 out out out out 9 425 out out out 1 50 out out out 10 425 out out out out 11 425 out out 1 00 out out out out out 11 800 out out out out out WinOpt HDR User Manual Solution Nr 31 Objective values Best Surface_Variance_PTV Objective value 0 029947 Best Volume_Variance_PTV Objective value 0 082574 Best D Dcrit 2 for urethra Objective value 0 000025 Best Objective value 0 000000 Implant Geometry Reconstruction File Contour File PTV Volume cm 3 52 7220 Surface cm 2 70 2718 Surface
13. 0 6707 Maximum Index 69 Maximum 1 5543 DDVH from sampling points used in optimization Dose D_ref counts 0 050000 0 150000 0 250000 0 350000 0 450000 0 550000 0 650000 0 750000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 20 270721 87 839789 0 850000 114 867417 0 950000 121 624324 1 050000 148 651951 1 150000 118 245870 1 250000 1 350000 1 450000 1 550000 1 650000 1 750000 1 850000 1 950000 2 050000 2 150000 9 450000 64 190615 23 649174 0 000000 3 378453 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 WinOpt HDR User Manual 9 550000 9 650000 9 750000 9 850000 9 950000 CDVH from sampling points used in optimization 0 000000 0 000000 0 000000 0 000000 0 000000 Dose D_ref percent 0 000000 0 100000 0 200000 0 300000 0 400000 0 500000 0 600000 0 700000 0 800000 0 900000 1 000000 1 100000 1 200000 1 300000 1 400000 1 500000 1 600000 1 700000 9 700000 9 800000 9 900000 100 000000 100 000000 100 000000 100 000000 100 000000 100 000000 100 000000 97 115385 84 615385 68 269231 50 961538 29 807692 12 980769 3 846154 0 480769 0 480769 0 000000 0 000000 0 000000 0 000000 0 000000 Statistics for PTV Volume of Sampling points 500 Mean 1 4490 Average Deviation 0 3173 Standard Deviation 0 4168 Variance 0 1737 Skewness 0 9043 Kurtosis 1 0601 Minimum Index 340 Minimum 0 6875 Maximum Index 161 Maximum 3 1165 DDVH f
14. 1 Dwell j yr SOUT PTV Critical Structure Fig 1 A single cylindrical shaped radioactive source of high activity e g 1 2Ir driven and controlled by a computer system is stepping within hollow catheters which have been previously implanted into the tumor The dose is calculated at a small number of points sampling points The contribution to the dose at each such point depends on the source dwell time the distance and the angle 9 Today the majority of treatment planning systems in brachytherapy such as Nucletrons PLATO system use phenomenological optimization methods such as geometrical optimization Additionally most of the algorithms used have the so called artificial problem of negative times which in principle does not exist and artificial methods such as setting the negative dwell times equal to zero and applying a dose renormalization are therefore used 20 50 of the dwell times that are always negative as a result of the optimization are arbitrarily set equal to 0 Dose normalization is then applied to rescale the dose at a specified number of dose points Another method used is to introduce constraints on the objective function such as gradient constraints between dwell times of adjacent dwells which reduces but does not completely avoids negative solutions Still up to 20 of the dwell times can be negative There is no reason WinOpt HDR User Manual 10 why a gradient based restriction should be applied to the obje
15. 100 00000 1 We sort the list of the 11 solutions according to the value of the coverage of the PTV with the prescription dose We find that the largest coverage is 94 2 for solution 151 0 86143 0 86307 0 87181 0 85697 0 88494 0 06242 0 88121 0 86186 0 85995 0 86262 0 85463 94 20500 92 54000 91 21000 90 65500 90 28000 90 26500 83 39000 89 09500 58 30500 67 23500 87 10500 WinOpt HDR User Manual 85 The isodose distribution is shown with the isodose of the prescription dose and the isodose of the critical dose for the urethra We have now a solution that satisfies to a great extent the OAR constraint The coverage of the PTV is larger than with the conventional method This is a consequence that the critical structures are considered in the optimization process and the dose normalization is absolute and not normalized to dose points on the PTV surface that imposes constraints on the obtained dose distributions WinOpt HDR User Manual 86 Example of template based manual Inverse Planning First we load VOls and the template coordinates We assume that the VOIS are given in a right handed coordinate system Before importing the necessary Charisma files we have to specify the coordinate system using the following dialog which appears after the Preplan with template is called in the FILEIO dialog Pmi pu r Fa a F S ag EE Coordinate system Lompa tty 5 The afterloader should be selected and th
16. Dose Points 208 Dose Points 500 Body Volume cm 3 1213 8017 Surface cm 2 638 8287 Dose Points 500 Rectum Volume cm 3 73 8731 Surface cm 2 105 6765 Dose Points 500 urethra Volume cm 3 2 3627 Surface cm 2 13 1406 Dose Points 500 Surrounding Tissue Dose Points 1000 Number of Catheters 16 Number of Dwells 270 Number of Active Dwells 94 Number of Surface Points 208 Sampling Points for CS or PTV Volumetry 50000 Sampling Points for Body Volumetry 100000 Sampling points 0 950000 mm from Catheter center ignored Dosimetric Kernel from Monte Carlo Lookup Table Source Strength 370 000000 GBq 40821 996772 U Prescription Dose 100 000000 cGy Here are the dwell times in seconds Dwell position weights in absolute units s Nr 1 Catheter 1 Dwell 1 Weight s 0 285665 x 89 900 y 60 219 z 94 800 Nr 2 Catheter 1 Dwell 3 Weight s 0 196752 x 89 900 y 60 219 z 84 800 Nr 3 Catheter 1 Dwell 4 Weight s 0 104952 x 89 900 y 60 219 z 79 800 Nr 4 Catheter 1 Dwell 5 Weight s 0 160361 x 89 900 y 60 219 z 74 800 Nr 5 Catheter 2 Dwell 1 Weight s 0 493317 x 99 900 y 60 301 z 92 300 Nr 6 Catheter 2 Dwell 2 Weight s 0 005404 x 99 900 y 60 301 z 87 300 Nr 92 Catheter 15 Dwell 3 Weight s 0 606145 x 114 771 y 90 425 z 68 550 Nr 93 Catheter 16 Dwell 1 Weight s 0 000310 x 99 750 y 95 301 z 74 800 Nr 94 Catheter 16 Dwell 2 Weigh
17. Imports VOIS Catheters and Loading from in Charisma format WinOpt HDR User Manual 37 The Dose Parameter Dialog This dialog is used to set the prescription dose and the source parameters Additionally the number of sampling points used in the optimization can be set Sets the prescription dose and the source strength required for the optimization Shows information about the VOIS and the sampling points Starts the auto activation dialog WinOpt HDR User Manual 132lr Nucletron Classic Flr Nucletron Classic 1921r Nucletran New 192Ir Mucletran PER 192 Varian Varsource 38 Used a dosimetric look up table for dose calculation Use for various sources a TG43 based interpolation function for the dosimetric kernel Use an invariant dosimetric kernel P i f HET aat Sa h Pane Geometry and samping This dialog can be used to specify the number of sampling points to be used in the VOIS for the anatomy based optimization Additionally the catheter margin can be modified or the number of points on the PTV surface per cm A default value of 3 Points cm is used and recommended Additionally the number of sampling points can be selected with which the final DVHs should be calculated with high accuracy Radius defines the outer catheter radius of the catheter used A margin can be used so that sampling points additionally inside this margin around the catheters are excluded in the dose calculation
18. NPICNICNVOI 1 1 Point O Begin Coordinates 123 25 23 12 12 34 Image 1 End 2272 227 107 WinOpt HDR User Manual Point NPICNICNVOI 1 1 1 Begin Coordinates 123 25 23 12 12 34 Image 34 End End End End 108 WinOpt HDR User Manual 109 3 The Catheters Charisma File Here is an example of the Catheters file which contains information about the Catheter type and geometry C mia sia sia mia ul ia mia mia ac all mia vl al sia al oa a a a vl a a a m al a vl a al MA al a al a al a al m a al al a al HF Reconstructed Catheters Data File CHARISMA R Vs 1 0 0 1 C opyrights Pi Medical Ltd All Rights protected ud sia mia mia ul a mia mia ac ul ha vlt al a al u a a a vl a a m al a vl a al m al a al a al al m a al al a al a E Catheter Describing Points X Y Z Image Status Image The number of the Image 0 N 1 with N the total number of Images 1 ifthe Catheter Describing Point doesn t lie on a Image All coordinates according to the World DICOM Coordinate System All dimensions are in mm Density values are given in g cm Date is given in mm dd yyyy Time is given in hh mm ss Mh mia sia ul a mia mia ac ul ia mia vl uk al a al u ac u a vl a a m a a vl a al a al EEE EEE EEE EEE al al E CHARISMA Software Version XX XX XX XX Reconstruction File Version XX XX XX XX File creation date and time mm dd yyyy hh mm ss KHRHHHHHHHHRHH HHH HHH H Patient Name SurName GivenName Patient ID XXXXXXXXXXXXXXX Num
19. and optimization WinOpt HDR User Manual 39 2 ONCE Fatameier This dialog is used to define the source strength or activity and the prescription dose These parameters have to be supplied for the use of the optimization algorithms The source is characterized by its strength in units of U or as activity in units of GBq or Ci The prescription dose is specified in cGy WinOpt HDR User Manual 40 The Template based Inverse Planning Dialog This dialog is used to set the parameters for pre planning It includes a template based catheter auto implantation possibility Sa Open and load a needle file that contains geometrical information of various needles lt Displays the list of all needles loaded A needle can be selected from this list WinOpt HDR User Manual 41 Provides an information about the loading of the catheters obtained by the automatic template based implantation algorithm Calls the algorithm to perform an automatic template based catheter implantation E ion i emplate D jas ed inverse Planning This dialog appears if a template based optimization should be performed and the VOIS and template position is loaded It contains information about the template and the catheters and afterloader parameters such as source step etc Sas Gf eeu i ite roson WinOpt HDR User Manual i empie Foaftamelers Offenbach Standard Totfenbach Standard This dialog is used for the selection of
20. geometric distribution the optimization will finally produce the best possible result An additional goal is to use the minimal number of dwell positions An analytic determination of the optimal distribution is currently not possible Therefore optimization methods are used which examine a subset of possible configuration out of which the best is selected These methods assume some importance factors with which the various objectives are combined An additional objective is a WinOpt HDR User Manual 32 minimum number of source dwell positions These objectives are of different scales and therefore some complex tuning of importance factors is used Only a single set of importance factors is usually used The result is that the optimization is limited in a very small part of the vast objective space and therefore the result could be of poor quality and the optimization has to be repeated with a different set of importance factors Additional the planner may think that the result is the best he could get but by limiting the search into a small set of the available space he she actually does not know what is actually possible The pre planning solution gives an optimal solution for the objective of the planner It requires that the catheters and source dwell positions are exactly known and fixed This is of course possible only to some extent WinOpt HDR offers the possibility to get the maximum possible out of the planners objectives given a distributio
21. list can be sorted in descending order for each of the objectives used in the optimization or the corresponding DVH values The smallest value for each such quantity corresponds to the best solution found in the optimization Additional constraints can be applied on the possible values thus reducing the number of solutions The full list without constraints shows what possible dose distributions can be realized It also shows the correlations and trade offs between the objectives The planner can select according to his her preferences a single non dominated solution Contrast to this interactive selection of a single solution some objective functions can be applied for the automatic selection of a single solution One must be of course cautious because a simple objective function which simulates the decision making process still does not exist in this or any other decision making tool One decision tool is based on the Conformal Index 1 corrected and extended now by D Baltas for the inclusion of critical structures proposed as a measure of implant quality and dose specification in brachytherapy which takes into account patient anatomy both of the tumor and COIN c c 2 normal tissues and organs COIN for a specific dose value D is defined as _ PTV PTV The coefficient is the fraction of the PTV PTV with dose values at least D 3 1 PTV C m 4 D The coefficient c is the fraction of the total calculated volume
22. sfr r sf 1 i i l In Eq 1 4 is the position of the i source and N the total number of sources fir r is the dosimetric kernel describing the dose rate per unit source strength at r from a source positioned at f s is proportional to the air kerma strength of the source For the case of HDR afterloader eth s is equal to 7 where 7 is the dwell time of the 1 source dwell position and is the air kerma strength of the single stepping source We use dose calculation point Look Up tables LUT in which the kernel values f for each dose calculation point and source dwell position pair is calculated and stored in the table once in a preprocessing step If we ignore the preprocessing time then the calcucation time for dose distributions is independent of the form of the dosimetric kernel This enables us to use realistic kernels obtained by sophisticated Monte Carlo simulation routines 11 The calculation requires only a file which contains the dosimetric LUT The dose distribution around a cylindrical source is not isotropic due to the attenuation of the photons in the active source material the encapsulation material the source drive cable etc Due to the cylindrical rotational symmetry the dose rate in a uniform isotropic medium is a function of r and only The orientation of the source is determined from the catheter geometry At each source dwell position a vector is calculated which is parallel to
23. the efficient calculation of DVHs of critical structures and the PTV a stratified sampling technique can be used 7 in which the regions also known as strata are defined by either optimal oriented bounding boxes or minimum enclosing spheres Fig 6 Contours of a critical structure myelon and the catheters of an implant showing the original bounding box white and the optimized bounding box red WinOpt HDR User Manual 17 Triangulation and Collision detection Anatomy based optimization assumes that the geometry of the PTV and the OARs is given WinOpt HDR assumes that the VOIS are given in form of contours defined by points on the constours obtained from parallel image slices from CT ultrasound or other imaging modalities We apply a triangulation of these points for the reconstruction of their surface This triangulation defines the geometry of the VOIS and is used for the calculation of the volume of the VOIS excluding parts of OARs in these VOIS such as the urethra inside the prostate case During the manual semiautomatic or automatic segmentation there may be some overlap of VOIS We apply a collision detection algorithm and report all triangles of all combination of VOIS which are in contact or if they intersect WinOpt HDR reconstructs the surface of the PTV Body and OARs by a triangulation VOIs with branches are not supported WinOpt HDR uses the triangulation algorithm of Fuchs et al to reconstruct the surface of the
24. the following conditions are called normalized importance factors n 1 gt Vj w 20 yw l Jel Two methods are supported for the generation of importance factors 1 Random distributed Importance Factors Importance factor vectors are generated with uniform probability p w for which YS CS p w dw p w dw V S V S wes where S is the set of all normalized weights and and S the subset of it V S and V S are the Euclidean hyper volume of S and S respectively The benefit of this method is that the Pareto front can be sampled with continuously resolution 2 Uniform distributed Importance Factors In this method each individual importance factor for each objective takes one of the following values l k 1 0 k where k is the sampling parameter n k For n objectives and a sampling parameter of k we have n such combinations While this method requires a precalculation of the imprtance factors Its benefit is that the distribution is uniform and that it avoids clusters and voids such as in the random distributed sampling case WinOpt HDR User Manual 29 Decision making tools We have implemented simple decision making tools that enables a planner to select a single solutions from the Pareto set based on objective values and DVHs Solutions can be selected from the list of all objectives and values of DVHs at the prescription dose for the Target and the critical structures respectively The
25. the optimal distribution of sources in space in an economic way i e finding the smallest possible number of source dwell positions without significantly reducing also the optimization quality in comparison to a solution a very large number within realistic limits of sources It should become clear soon or later that only multiobjective optimization should be used for the inverse planning problem The new optimization methods do not require normalization on some arbitrary set of dose points They do not require arbitrary correction mechanisms They do not require arbitrary importance factors All these constraints in the past have due to ignorance or due to limiting computational possibilities introduced restrictions on the optimization result The dominance of the market by a single provider for treatment planning systems in brachytherapy has helped to establish the old methods and it is now difficult to replace these by the new methods It is true that the new optimization methods require a slightly more complex cooperation with planners and still the decision making process of these multiobjective methods can be improved with some additional tools A dose optimization procedure which does not require any human decision making process is difficult if not impossible to be realized WinOpt HDR User Manual 13 Dose computation using dosimetric Look Up Tables The dose D t at r x y 2 is conventionally calculated using Eq 1 Ns Ns Dlr
26. 0 0 00000 1 00000 Inac Dwell 7 Pos 89 9002 60 2186 64 8000 t 0 00000 0 00000 1 00000 Inac Dwell 8 Pos 89 9002 60 2186 59 8000 t 0 00000 0 00000 1 00000 116 WinOpt HDR User Manual 117 Inac Dwell 16 Pos 89 9002 60 2186 19 8000 t 0 00000 0 00000 1 00000 Inac Dwell 17 Pos 89 9002 60 2186 14 8000 t 0 00000 0 00000 1 00000 Catheter 2 List of Describing points Number of Describing points for this catheter 2 Point 1 Pos 94 9000 60 2599 9 8000 Point 2 Pos 94 9000 60 2599 102 0500 Tot Source Nr 0 in Catheter Nr 1 at dwell 1 Tot Source Nr 1 in Catheter Nr 1 at dwell 3 Tot Source Nr 2 in Catheter Nr 1 at dwell 4 Tot Source Nr 3 in Catheter Nr 1 at dwell 5 Tot Source Nr 4 in Catheter Nr 2 at dwell 1 Tot Source Nr 5 in Catheter Nr 2 at dwell 2 Tot Source Nr 6 in Catheter Nr 2 at dwell 3 Tot Source Nr 7 in Catheter Nr 2 at dwell 5 Tot Source Nr 8 in Catheter Nr 2 at dwell 6 Tot Source Nr 9 in Catheter Nr 3 at dwell 1 Tot Source Nr 10 in Catheter Nr 3 at dwell 2 Number of Active Sources found 94 Act Dwell Nr 1 Catheter Nr 1 Dwell Nr 1 Pos 89 900166 60 218575 94 800000 Act Dwell Nr 2 Catheter Nr 1 Dwell Nr 3 Pos 89 900166 60 218575 84 800000 Act Dwell Nr 3 Catheter Nr 1 Dwell Nr 4 Pos 89 900166 60 218575 79 800000 Act Dwell Nr 4 Catheter Nr 1 Dwell Nr 5 Pos 89 900166 60 218575 74 800000 Act Dwell Nr 5 Catheter Nr 2 Dwell Nr 1 Pos 99 899825 60 301217 92 300000 Here log of optimization run WinOpt HDR v 2 00 New Optimi
27. End Volume 123 25 23 12 12 34 1234567890 123 Surface 1234567890 123 Inner Surface 1234567890 123 Number of Contours NCO Contour 0 Begin Type XXXXXX e g not on an Image or 2 Number of Points Point 0 Begin End NPCOO Coordinates 123 25 23 12 12 34 Image 1 WinOpt HDR User Manual Point NPC00 1 Begin Coordinates 123 25 23 12 12 34 Image 34 End End Contour NCO 1 Begin Type XXXXXX e g not on an Image or Number of Points NPCNCOO 1 Point O Begin Coordinates 123 25 23 12 12 34 Image 1 End Point NPCNCOO 1 1 Begin Coordinates 123 25 23 12 12 34 Image 34 End End Number of Inner Contours NICO Contour 0 Begin Type XXXXXX e g not on an Image or Number of Points NPICOO Point O Begin Coordinates 123 25 23 12 12 34 Image 1 End Point NPICOO 1 Begin Coordinates 123 25 23 12 12 34 Image 34 End End 227 227 104 WinOpt HDR User Manual 105 Contour NICO 1 Begin Type XXXXXX e g not on an Image or 77 Number of Points NPICNICOO Point 0 Begin Coordinates 123 25 23 12 12 34 Image 1 End Point NPICNICOO 1 1 Begin Coordinates 123 25 23 12 12 34 Image 34 End End End VOI NVOI 1 Begin Name BBBBBBBBBBBBBB Type Volume Type Compact or Hole Wall Thickness XXXX XXX Extraction Begin Reference VOI Begin VOI No XXX Name CCCCCCCCCCCCC End Type Type of Extractio
28. HH HH HH HH HH EEE EEE EE EHH Wall Thickness If Volume Type compact then wall thickness 0 0 mm If Volume Type compact then Number of Inner Contours NIC 0 If Volume Type compact then Inner Surface Surface FEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE EEE EEE EEE HEH H All coordinates according to the World DICOM Coordinate System All dimensions are in mm All surfaces are in cm All volumes are in cm Date is given in mm dd yyyy Time is given in hh mm ss PEER EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE EEE EEE EEE EE EHH CHARISMA Software Version XX XX XX XX Reconstruction File Version XX XX XX XX File creation date and time mm dd yyyy hh mm ss KHHHHHHHHH HHH HHP HHH HH HHH HHH Patient Name SurName GivenName Patient ID XXXXXXXXXXXXXXX Number of VOIS NVOI VOI Data Begin VOI 0 Begin Name BBBBBBBBBBBBBB Type CS BODY or PTV Volume Type Compact or Hole Wall Thickness XXXX XX Extraction Begin Reference VOI Begin VOI No XXX Name CCCCCCCCCCCCC WinOpt HDR User Manual End Type Method Begin End End Connections Begin Type of Extraction e g 3D volume 2D on CT plane Type Isotropic or Ellipsoid or Margins Begin XXX XX XXX XX XXX XX XXX XX XXX XX XXX XX End 103 Here enter information for connection amp connection type between two End BBOX Begin Origin 123 25 23 12 12 34 Vector 1 123 25 23 12 12 34 Vector 2 123 25 23 12 12 34 Vector 3
29. Kostas Karouzakis Stavroula Giannouli and Natasa Milickovic USER S MANUAL Email mailto mlahanas gmx net Web http www mlahanas de WinOpt HDR User Manual 2 Information in this document is subject to change without notice This User s Manual can be reproduced stored in a retrieval system or transmitted in any form or by any means electronic or mechanical including photocopying and recording for any purpose The following trade names are referenced throughout this manual Windows NT Windows 2000 Windows 95 98 are trademarks of Microsoft Corporation Pentium Pentium II Pentium Ill and Pentium V are trademarks of Intel Corp WinOpt HDR User Manual 3 INHOAUCH ON niet u sur verrttrn wrortenter trent trier 8 Dose computation using dosimetric Look Up Tables uussessersorssnssossensonene 13 Sampune Points nasien aha 14 Opumizcd Doundino DOXE S eiii eer Teeny er mercer nee ere er errr ery 16 Triangulation and Collision detection urssussosssnssonssnssnnssnnsonsnnsnnsnnnsnnnsnnsnnnsnne 17 Dose Optimization in HDR brachytherapy cccrcccccrscccrsscccsssccssscccssscccsssccssssccssssccssscecsssceeess 18 Dose optimization USING variances scesc eccsccoscesscencessesscsssessooncesecosnscdavsnsoeessconseees 18 Dose optimization using dose volume histogram based objectives 21 Single objective weighted sum optimization eurssorssonsssonssonssonssnnnssnnsnonssnnnenene 23 True multiobjective Optimization
30. Status Inactive Weight 0 000000 End Catheter N 1 Begin Number of Points NSN 1 Point O Begin Coordinates 145 15 234 38 223 00 Status Inactive Weight 0 000000 End Point 1 Begin Coordinates 145 15 234 38 223 00 Status Active Weight 0 123456 Point NSN 1 1 Begin Coordinates 144 60 234 46 218 00 WinOpt HDR User Manual Status Inactive Weight 0 000000 End End End 114 WinOpt HDR User Manual 115 The WinOpt HDR Log File A logfile which stores information during the running of WinOpt HDR is created and can be found as WinOpt_Log txt in the working directory Here is a example WinOpt HDR v 2 00 Anatomy based HDR 3D Dose Optimization WinOpt HDR by Michael Lahanas Email mlahanas gmx de www www mlahanas de Developed by Michael Lahanas Kostas Karouzakis Stavroula Giannouli Maria Papagiannopoulou and Natasa Milickovic from Medical Physics and Engineering Department in Offenbach Germany under the supervision help and support from Prof Dimos Baltas Monte Carlo Dosimetric Look up Table from Pantelis Karaiskos et al University of Athens OS time 11 40 04 OS date 10 22 01 Monday 22 October 2001 WinOpt Log File Check Topology of PTV Contours Polygon is not simple crossing of edges 39 900002 241 300003 40 849998 231 800003 and Crossing is reported 38 950001 235 600006 40 849998 236 550003 Less than 0 01 mm between contour points Points with distance less than 0 01 m
31. The normalization at specific points introduces a constraint that produces practical always points in the PTV with dose values below the prescription dose cold spots Also it cannot consider critical structures inside the PTV or in its vicinity In order to avoid to some extend under dosage or dose values above the critical dose in the critical structure it is necessary to rescale the dwell times of the obtained solution i e to use some other normalization factor This problem is always present more or less since it is impossible to satisfy all conflicting objectives simultaneously It is a consequence of the limited number of sources Additional the source characteristics are such that the possibilities of the obtained dose distribution are limited In some cases the planner is willing to sacrifice then overdose of a portion of a region at risk in order to improve the probability of curing the disease Constraints could be no more than x of this region can exceed a dose value of Da One advantage of these so called partial volume constraints is that the results in terms of objective values are more intuitive to understand We used the following set of objectives formulated as partial volume constraints 1 Fraction of PTV including surface with D lt D 2 Fraction of PTV including surface with D gt D 3 Average squared dose in the surrounding normal tissue lt D gt 4 Fraction of CS with D gt D D is the prescription dos
32. VOIS from the points on the parallel slices WinOpt HDR User Manual 18 Dose Optimization in HDR brachytherapy Dose optimization using variances One solution of conformal HDR brachytherapy is to obtain a dose distribution such that the isodose of the prescription dose coincides with the PTV surface In principle with this approach the use of an additional objective for the surrounding body is not necessary This objective can be expressed as the problem to determine the time of the source dwells such that the resulting dose variance f of the sampling points dose points uniformly distributed on the PTV surface is as small as possible In order to avoid excessive high dose values inside the PTV we require a small as possible dose distribution variance f inside the PTV Due to the source characteristics these two objectives are competing We use normalized variances for the two objectives 1 amp D my f Y D m s me i s ty aN j y Where m and is the average dose value on the PTV surface and in the PTV volume and N N the corresponding number of sampling points For these objectives the Pareto tradeoff surface described later in the text is convex and gradient based algorithms converge to the global Pareto front 13 15 In the past dose points have been use limited on the target contours With this approach increasing the number of sampling points does not increase the coverage of the PTV We develo
33. YHETO00 oFrectum DWH Z DOD of Urethra Save List 0 90399 95 38000 0 00500 49 11500 Delete Constraints 0 98977 34 18500 0 02000 43 089500 0 88797 33 41500 0 02000 30 09500 0 86596 92 69500 0 02500 12 23000 0 84934 92 02000 0 02000 5 53000 Hinne 0 83697 31 56500 0 04500 2 12500 0 89330 30 933000 0 06500 0 84500 0 82895 90 47500 0 08000 0 19000 0 82118 90 05000 0 10500 0 06000 EPO gt 0 81490 89 70500 0 12500 0 00500 0 80405 89 32000 0 15500 0 00000 0 79884 89 02000 0 18000 0 00000 080014 88 67500 0 21000 0 00000 0 78811 88 423000 0 22000 0 00000 0 78998 88 11000 0 22000 0 00000 0 77595 87 74500 0 23500 o po000 0 76082 87 320000 0 24000 O 00000 0 76081 86 88500 0 26500 0 00000 0 75900 86 36500 0 26500 0 0000 ul 1 ii 3 4 5 G T g 3 Abbrechen We observe that the there is a trade off between the part of the urethra that receives a dose above the critical dose value and the PTV coverage We see as expected that the best conformity index is achieved for the solution that considers only the variance on the PTV surface For this solution although 50 of the urethra receives a dose above the critical dose The coverage for the PTV is 95 4 The dose distribution is shown in the following figure Dose distribution for Solution Nr 0 in the Filter Dialog smallest PTV surface dose variance We select solution 6 We see that less than 1 of the urethra receives a dose larger than the critical dose value No
34. a needle type from a list of needles previously imported 42 WinOpt HDR User Manual 43 Autoactivation This dialog is used for the auto activation algorithm The PTV and Organs at risk are listed and the minimum distance of the source dwell positions from the corresponding VOI in mm It is used to select only source dwell positions that are at a distance to a corresponding VOI larger than a specified value WinOpt HDR User Manual 44 Windpt HOR Result of the algorithm which determines which catheter can be used in the template based inverse planning for a prostate implant where the rectum and the urethra is also shown Catheters that through their path hit the urethra are not selected WinOpt HDR User Manual 45 Template Catheter Loading This dialog shows which catheters can be used in a template based preplanning 46 WinOpt HDR User Manual ACTIVE NEEDLE MAP 1 1 LXK XK XK 1 gt X X X X 1 1 gt K K XK X X X L K X X rot 1 1 XK KKK X KKK 1 KK KKK XXK XK 1 1 KKK KK XK KKK KKK KX XK KKK KK KK X 1 KKK KKK This table shows marked with an X the grid position on the template where a catheter can be inserted ACTIVE DWELL MAP LOOOoooooooo0o00o0000o0 ooooooooN oo O O Wooococo cnn0000o0 sooonnnp9oon9 ry nn N oO oO QDAoorwMounvtrnonunwtrtnNnoond OOOO LOLO 4 O O DD DD OD FTO OO Ooocoo ovvownwwwaWawoo SOoonr 2 2 DD DOM TOOO Dooocovooocoooo soooo
35. ameters If this is so then there exist many in principle infinite solutions A multiobjective algorithm does not provide a single solution but a representative set of all possible solutions Out of these representative solutions a single final solution has to be selected It is a complex problem to automatically select such a solution and such methods have been proposed but then a planner would not know what alternatives solutions could instead be selected In problems where different sets of objectives have to be compared this information is valuable since it shows the possibilities a planner has for each such set WinOpt HDR User Manual 25 Multiobjective optimization Introduction and Definitions The multiobjective optimization MO problem also called multicriteria optimization or vector optimization can be defined as the problem of determining A vector of decision variables which satisfies constraints and optimizes a vector function whose elements represent the objective functions These functions form a mathematical description of performance criteria that are usually in conflict with each other Hence the term optimize means finding such a solution which would give the values of all objective functions acceptable to the designer We call decision variables x j 1 2 N for which values are to be chosen in an optimization problem In order to know how good a certain solution is we need to have some criteria for evaluatio
36. an y it is said that y is dominated by x or y is inferior to x We say that a point x in X is Pareto optimal if and only if there is no x e X for which f x dominates f x i e there is no x such that for k objectives V Le dha Ks IS Ste adi e 1 Kl EI TEN Each element in the Pareto optimal set constitutes a non inferior solution to the MO problem The problem has usually no unique perfect solution but a set of equally efficient or non inferior alternative solutions known as the Pareto optimal set Each point in this set is optimal WinOpt HDR User Manual 26 in the sense that no improvement can be achieved in one vector component that does not lead to a degradation in at least one of the remaining components That is there are no other solutions superior in all attributes The set of non dominated solutions lie on a surface on the boundary of the feasible and non feasible objective space known as the Pareto optimal front The Pareto set of the entire feasible space is called the global or true Pareto set In most cases there will be several optimal solutions in the Pareto sense and we have to look to the values of the objective functions in order to decide which values seems the most appropriate This process in which a solution is selected is called the decision making process In comparison to single objective algorithms which provide one single solution the task of multiobjective algorithms is to provide a representative sa
37. ber of Catheters N Catheter Data Begin Catheter 0 Begin Name aaaaaaa Material aaaaaaa Density XX XXX Outer Diameter XX XXX Inner Diameter XX XXX Length XXXX XXX min Free Length XXXX XXX Distance Tip 1st Source Position XX XXX Channel Length WinOpt HDR User Manual XXXX XXXK Distance 1st Reconstructed Point Tip XXXX XXXK Reconstructed Length XXXX XXXK Free Length XXXX XXXK Catheter N 1 Begin End End Name aaaaaaa Material aaaaaaa Density XX XXX Outer Diameter XX XXX Inner Diameter XX XXX Length XXXX XXX min Free Length XXXX XXX Distance Tip 1st Source Position XX XXX Channel Length XXXX XXX Distance 1st Reconstructed Point Tip XXXX XXX Reconstructed Length XXXX XXX Free Length XXXX XXX Catheter Describing Points Begin Catheter 0 Begin Number of Points NPO Point O Begin Coordinates 146 18 234 14 228 00 Image 0 End Point 1 Begin Coordinates 145 15 234 38 223 00 Image 1 110 WinOpt HDR User Manual 111 Point NPO 1 Begin Coordinates 127 67 224 33 83 00 Image 29 End End Catheter N 1 Begin Number of Points NPN 1 Point O Begin Coordinates 146 18 234 14 228 00 Image 0 End Point 1 Begin Coordinates 145 15 234 38 223 00 Image 1 End Point NPN 1 1 Begin Coordinates 127 67 224 33 83 00 Image 29 End End End WinOpt HDR User Manual 112 4 The Loading Charisma File Here is an example of the Loading file which con
38. c c ge at i V OAR 6 WinOpt HDR User Manual 31 Template based Inverse Planning Two are the main questions in HDR brachytherapy dose optimization a How and how many catheters should be inserted and which possible source dwells positions should be used b What is the duration of a radioactive source at each of these so called active source dwell positions such that the resulting dose distribution satisfies various objectives WinOpt HDR supports template based pre planning optimization and post plan optimization with any catheter topology An automatic template based catheter implantation routine is used The algorithm determines given a template from the anatomical structures PTV and OARs which catheters and to what extent they should be inserted given a minimum distance to the PTV surface and organs at risk This gives a partial answer to question a A template based inverse planning algorithm is supported i e given a template find the minimum number of catheters and source dwell times so that given criteria are satisfied Source dwell position PTV z Catheter Template Fig 11 Output of the automatic template based catheter implantation algorithm For each catheter find the free length of the catheter and the possible source dwell positions considering organs at risk It is therefore desirable to find also the position of the dwell sources such that using this
39. catheter reconstruction volumetric rendering image fusion etc It assumes that these possibilities exist in some other program and that the information can be exported so that they can be used by WinOpt HDR The methods presented in this program should be easily ported to any brachytherapy treatment planning system WinOpt HDR includes the necessary tools for data mining analysis and optimization with respect to multiple objectives WinOpt HDR has been found in comparison to Nucletrons planning system PLATO to produce superior and in the worst case compatible results Systems like PLATO provide only a single solution and or organs at risk cannot be considered except manually and if only using some phenomenological approximations WinOpt HDR offers the possibility to provide in principle all possible solutions It offers the planner the tools to select the best solutions out of this set Plato and other systems often use artificial methods to reduce negative dwell position weights These methods include constraints on the objective function which reduces but does not eliminate these solutions An artificial correction mechanism is applied by setting all negative dwell times equal to 0 This mechanism necessarily reduces the quality of the obtained solution WinOpt HDR avoids the generation of such non physical solutions It does not require any artificial constraint for the objective method and also no correction at any stage of the optimization proces
40. ctive function only to reduce negative times which can be avoided by a more natural and extremely simple method without imposing any restriction on the optimization result Other approaches use a constraint of positive dwell times They do not require some rather arbitrary modification of the objective function This in principle increases the number of optimization parameters by a factor of two If the constraints are satisfied only partial and negative dwell times are still present in the final solution again a correction mechanism must be applied An additional constraint applied on the optimization result is the use of a dose point normalization based on the dose of some points for example on the minimum peripheral dose or the average dose on the PTV surface This is because the definition of what conformal brachytherapy means is not mathematical well defined If the most important objective is to have a dose at least above some value in the entire PTV then a normalization based on the minimum peripheral dose can be used By using this normalization method it is true that the dose in the PTV will be everywhere high enough but how is then the dose distribution in the surrounding tissue and in organs at risk This method based on the normalization on a single point usually increases the dose in the surrounding tissue by in principle expanding the reference isodose so that it passes through this dose point at which the minimum dose was measured and if the
41. ctives 3 Optimization Objectives 1 PTV Surface Variance 2 PTV Volume Variance 3 Volume Variance D gt Dcrit of urethra Using WinOpt HDR Very Fast Deterministic Optimization Algorithm Include OAR urethra with 500 Sampling Points and Dcrit Dref 1 500000 Implant Geometry Reconstruction File Contour File PTV Volume cm 3 52 7220 Surface cm 2 70 2718 Surface Dose Points 208 Dose Points 500 Body Volume cm 3 1213 8017 Surface cm 2 638 8287 Dose Points 500 Rectum Volume cm 3 73 8731 Surface cm 2 105 6765 Dose Points 500 urethra Volume cm 3 2 3627 Surface cm 2 13 1406 Dose Points 500 Surrounding Tissue Dose Points 1000 Number of Catheters 16 Number of Dwells 270 Number of Active Dwells 94 Number of Surface Points 208 Sampling Points for CS or PTV Volumetry 50000 Sampling Points for Body Volumetry 100000 Sampling points 0 950000 mm from Catheter center ignored 118 WinOpt HDR User Manual 119 WinOpt HDR User Manual 120 The WinOpt HDR Solutions File The result of the optimization by WinOpt HDR can be found as WinOpt_Solution txt in the working directory It contains the solution which was saved in the Filter Dialog It is overwritten each time if a new solution is selected It contains information about the catheters source dwell positions the dwell times finally found relative and in absolute units Also it contains dose distributions in terms
42. ctors or weights the Pareto set is sampled randomly if it is convex and the algorithm is not trapped by local minima Random distributed importance factors are generated Specify number of solutions if random weights are selected WinOpt HDR User Manual 59 Displays for the PTV variance based single dose optimization method the objective and the corresponding importance factor Displays for the DVH based single dose optimization method the objective and the corresponding importance factor WinOpt HDR User Manual 60 This dialog is used for the setting of the parameters of the fast simulated annealing optimization algorithm WinOpt HDR User Manual 61 The Deterministic Optimization Dialoo Deterministic Post Plan Dose Optimization IE This dialog is used for the optimization of the surface and volume variance A single solution can be obtained The user must supply the importance factor for the volume variance The two dimensional Pareto trade off surface can be samples by the multiobjective weighted sum approach In this case the user has to supply the number of intervals to be sampled for each importance factor Single optimization run either surface variance only or if volume variance should be included then the user must supply the corresponding volume importance factor Multiple optimization runs the volume variance importance factor is sampled in n steps specified by the user There are t
43. deoff between the objectives WinOpt HDR offers a true multi objective inverse plan optimization method for template based implants In this method the optimization algorithm determines an optimal minimal subset of catheters that should be used without a significant loss in optimization quality This optimization is very complex and deterministic algorithms are very slow or are trapped in local minima due to the very complex objective function dependence on the dwell position weights and the catheter configuration Since the combinatorial complexity is extremely large a supported evolutionary multiobjective optimization algorithm is used which can handle this very complex optimization problem due to its so called implicit parallelism The user is supposed to supply a range of number of catheters to be used in the optimization WinOpt HDR User Manual 54 The Evolutionarv Optimization Dialog mu Itiobje ctive Evolution Post Plan 0 ptimizatio n po en This dialog is used to select the parameters and the objectives for a true multiobjective dose optimization method without the use of importance factors Two methods are supported PTV based using variances on the PTV surface and in the volume DVH based which is based on DVH based objectives For this method additional a constraint can be set to select only solutions with a fraction above the Prescription dose less than a specified value Up to three critical structu
44. e or lower dose limit and D is the high dose limit in the PTV lt D gt is the mean quadratic dose in the surrounding normal tissue and D is the critical dose for a critical structure organ at risk In this model we have two objectives for the target 1 2 one objective for the surrounding normal tissue 3 and for each organ at risk an additional objective 4 The inclusion of points on the surface for the objective 1 2 improves only the definition of the boundaries of the PTV The objectives 1 2 4 are strongly correlated with the DVH values at the corresponding dose values As a fraction they are normalized in the range 0 1 The dose values are expressed as fractions of the prescription dose If the sources are limited inside the PTV using the auto activation algorithm then the range of the values of the objective 3 is of comparable magnitude as the other objectives The use of the square of the dose value ensures that high dose values are more likely to be avoided than more uniformly distributed moderate dose values Fig 8 shows an ideal conformal dose distribution for the PTV which is a simple delta function at the prescribed dose value The corresponding ideal and a realistic cumulative DVH is show in Fig 9 WinOpt HDR User Manual 22 dV dD of PTV for optimal dose distribution A S 20 dV dD of real dose distribution xe 10 0 0 1 pp 2 3 4 Fig 8 Dose distributions are described by dose volume histogra
45. e because objectives are either cooperating or in conflict or indifferent WinOpt HDR User Manual 24 True multiobjective optimization Conventional dose optimization algorithms are single objective i e they provide a single solution This solution is found by a trial and error search method by modifying importance factors of a weighted sum of objectives This problem has been addressed currently and some methods have been proposed to find an optimal set of importance factors 8 9 Conventional optimization methods combine the target objectives and the objectives for the surrounding healthy tissue and of critical structures into a single weighted objective function The weight or importance factor for each objective must be supplied The obtained solution depends on the value of importance factors used One goal of a treatment planning system is the ability to assist the clinician in obtaining good plans on the fly Also it should provide all the information of the possibilities given the objectives of the treatment In order to explore the feasible region of the solution space with respect to each objective different values for the importance factors in the aggregate objective function must be given and then the results assed Furthermore the appropriate values of these importance factors differ from clinical case to clinical case This implies that for any new clinical case a lot of effort is necessary for their determination While current opt
46. e loading i e minimum and maximum number of dwells per catheter The Autoactivation parameters should be set WinOpt HDR User Manual 87 Autoactivation WO Distance mar PTy goog Rectum ie urethra 1 0000 x This dialog specifies the autoactivation parameters 1 e minimum distance of all dwells to each VOL WinOpt HDR After this the catheters which satisfy these constraints are selected and inserted automatically inside the PTV WinOpt HDR User Manual 88 ne Be m Be Ao 2 a cn Pi y a ee tcae E amp E oc u etetett k i 4 Po oP Be ho Eo oe l In the template view the selected catheters are shown The plane along the template normal can be moved with the z slider The intersection of the VOls is shown in the template view We move it at a height where the PTV cross section is maximal PTV Bird Template Catheters 5 5 5 5 baie Bee sail Catheter density 4 5 4 5 4 0 4 0 come 4 5 eama i Catheter Mr 3 0 3 0 Tan al 25 2 5 EFT 2 0 2 0 jE 1 5 1 0 1 0 We reduce the number of catheters by selecting only those which are inside every second template grid point WinOpt HDR User Manual 89 e e M M de A A om co mW Oo A A A A A AA A Era 5 0 4 5 4 0 ae 3 0 a 2 0 Tesa 1 0 By using the left mouse button double click we can manually switch catheters on and of
47. eighted sum to be optimized are w and w According to their values their vector sum defines a direction Fig 10 Optimal solution found for a combination of importance factors In the objective space the weighted sum is given by y w fi x w f x This can be written as f x f x gt The minimization of the weighted sum can be interpreted gt gt as finding the value of y for which the line with slope w w just touches the boundary of F It is therefore not possible to obtain solutions on non convex parts of the Pareto front with this approach We support two multiobjective optimization methods A multiobjective weighted sum method based on the fast simulated annealing algorithm 19 and a true importance factors free multiobjective optimization algorithm that uses a multiobjective evolutionary algorithm The weighted sum approach using the fast simulated annealing algorithm can practically provide only a small number of non dominated solutions Additional only convex parts of the Pareto tradeoff surface can be found It requires a large number of iterations but converges most likely to the global Pareto front The evolutionary algorithm uses a population that is guided towards the global Pareto front It provides 100 and more solutions in less than 10 minutes WinOpt HDR User Manual 28 Generation of uniform and random importance factors for multiobjective optimization Importance factor vectors w J l n that meet
48. eport 2000 12 M Lahanas D Baltas and N Zamboglou Anatomy based three dimensional dose optimization in brachytherapy using multiobjective genetic algorithms Med Phys 26 1904 1918 1999 Abstract Preprint in pdf format WinOpt HDR User Manual 97 13 M Lahanas N Mihckovic M Papagiannopoulou K Karouzakis D Baltas and N Zamboglon Application of a Hybrid NSGA II Multiobjective Algorithm for Anatomy based Dose Optimization in Brachytherapy EUROGEN 2001 Evolutionary Methods for Design Optimisation and Control with Applications to Industrial Problems Athens Greece 19 21 September 2001 Abstract Preprint in pdf format 14 R Van der Larsen T P E Prins Introduction to HDR brachytherapy optimization In R F Mould J J Batterman A A Martinez and B L Speiser eds Brachytherapy from Radium to Optimization Veenendaal The Netherlands Nucletron International pp 331 351 1994 15 N Mehckovic M Lahanas M Papagiannopoulou K Karonzakis D Baltas and N Zamboglon Application of a Multiobjective Genetic Algorithms in Anatomy based Dose Optimization in Brachytherapy and its Comparison with Deterministic Algorithms EUROGEN 2001 Evolutionary Methods for Design Optimisation and Control with Applications to Industrial Problems Athens Greece 19 21 September 2001 Abstract Preprint in pdf format 16 N Milickovic S Giannouli D Baltas M Lahanas C Kolotas N Zamboglou and N Uz
49. ers or actions of the corresponding Menu of the Main Toolbar E The Progress bar appears when a lengthy calculation starts showing percent of the calculation performed sADKOaSEE alt WinOpt HDR User Manual 35 Main Toolbar used to select the corresponding module and dialog in the right side of the user interface Import Module import VOIS Catheters and Loading in Charisma format Dose Optimization Module Settings for the display of two and three dimensional isodose lines and surfaces O A Exit Stops the WinOpt HDR application an m zl Export Loading in Charisma Format Documentation of Optimization results E 6 Er Parameter Setting such as prescribed dose source strength and sampling points ZE Volumetry module Display of volume and surface areas of VOIS posam B Generates a compressed image of the Anatomy Window in TIFF format En Opens or closes the Bi Objective Window that shows the Population of the multiobjective evolutionary algorithm Toggle between solid and wire frame display format for the VOIS in the anatomy window Starts and stops rotation of implant and VOIS around the z Axis Bie o Prints the Image Window WinOpt HDR User Manual 36 The FILE I O Dialog FILE I O This dialog is used to for the import of the VOIS Catheters and Loading File in Charisma format which define the geometry and the source distribution necessary for the anatomy based dose optimization
50. es the number of points each importance factor has to be sampled for each objective If we divide the range of each importance factor in n intervals then for k 2 2 objectives the number of combinations is of the order nk k 1 k D n k 1 For three objectives this means that the following set of importance factors will be used WinOpt HDR User Manual os o fos 0 57 WinOpt HDR User Manual 58 Example of the importance factors used for k 3 objectives divided in n 3 steps For all six cases we have witw2tw3 1 Apply a PTV variance based dose optimization using variances Apply a DVH based dose optimization Only a single run with importance factors set by the user Multiple runs with importance factors determined by the number of intervals for each importance factor Uniform Weights With uniform distributed importance factors or weights the Pareto set is sampled uniformly if it is convex and the algorithm is not trapped by local minima Uniformly distributed importance factors weights are generated Specify number of importance factors objective Please consider that number of final solutions increases exponential with number of objectives Eq 7 and only maximal 500 solutions are currently supported Reports the total number of solutions Win pt HDR W2 00 Example of the output for 3 Objectives and 5 steps per objective Random Weights With random distributed importance fa
51. f For example we select catheters which are on the periphery and some around the PTV center Finally we press the ACCEPT button in the template view WinOpt HDR User Manual 90 Windpt HOR Template based Multiobective Evolutionary li Settings i Organs at Risk DAR Rectum urethra Consider DAR ar Objectives as Ecrit 11 5000 Drel 1 5000 Dret 11 5000 Dref Constraint E Objectives PTY Low PTW High Constraints Fir Rufe PtViow lt 030000 Number of Catheters Some other selection criteria can be realized but this is not a true inverse planning An automatic optimal selection of catheters can be realized by using the template based multiobjective evolutionary inverse plan algorithm but this method is not available in the DEMO version WinOpt HDR User Manual 91 WindptHOIR Our manual selection results in the following distribution of source dwell positions We assume that the catheters are optimal distributed and then we can use the post plan optimization algorithms In the true multiobjective optimization also a optimization of dwell times is included We use the deterministic algorithms and specify a critical dose of 1 5times the prescription dose We consider 40 runs with random distributed importance factors in order to cover a part of the Pareto front We sort the solutions according to the COIN value without OARs We see that the maximum conformity still requires that 50 of the uret
52. he PTV surface defined by 208 dose points in this example Number of Dose Points Sum of relative weights 20 458434 Average relative weight 0 217643 Reference dose value Mean dose value on PTV surface Normalization Factor 1 000000 Dose Scaling Factor 1 000000 208 WinOpt HDR User Manual 124 High Statistics with 20000 Sampling points DDVH Differential Dose Volume Histogram for the PTV High Statistics of DDVH for the PTV 0 050000 0 150000 0 250000 0 350000 0 450000 0 550000 0 650000 0 750000 0 850000 0 950000 1 050000 1 150000 1 250000 1 350000 1 450000 1 550000 1 650000 1 750000 1 850000 1 950000 2 050000 2 150000 2 250000 2 350000 9 250000 9 350000 9 450000 9 550000 9 650000 9 750000 9 850000 9 950000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 394505 8 758019 20 540579 33 559255 44 342402 53 968332 56 940273 56 072361 54 047233 51 390897 34 900574 24 012226 16 174719 13 123878 8 731718 6 548789 5 996481 4 602562 0 026300 0 052601 0 105201 0 026300 0 052601 0 026300 0 026300 0 000000 CDVH Cumulative Dose Volume Histogram High Statistics of CDVH for the PTV 0 000000 100 000000 0 100000 100 000000 0 200000 100 000000 0 300000 100 000000 0 400000 100 000000 0 500000 100 000000 0 600000 100 000000 0 700000 0 800000 0 900000 1 000000 1 100000 1 200000 1 300000 1 400000 1 500000 1 600000 1 700000 1 800000 1 900000 2 000000 2 100000
53. hen the powerful gradient optimization method converges according to the Kuhn Tucker theorem for convex or quasi convex functions to a global solution No local minima are present The gradients guide the algorithm fast to the global minimum defined by a minimum variance on the PTV surface In this approach the objective function f is scale invariant i e if the source dwell times are t t tn then f t t ty flat atz aty where a is any positive a gt 0 number A benefit of this method is that the search space can limited to any range such as 0 1 For other objective functions that are not scale invariant an absolute scale must be given The optimization method requires that only source dwell positions inside the target should be used If source dwell positions outside the PTV are included then high dose values outside the PTV may exist It is therefore necessary to use the auto activation algorithm The objectives in terms of variances allow the use of gradient based algorithms that converge very fast to the global minimum A single optimization requires only a few seconds Dose Optimization sampling Points on the PTV surface Fig 7 Example of an optimization for the variance based method The PTV catheters and source dwell positions are show Points on the PTV surface are included at which the dose variance around the prescription should be minimized The result is show as the isodose surface of the prescription dose
54. his is called permanent implantation and for this technique low activity and low energy radioactive sources radioactive seeds are used This method is call LDR method e A single radioactive source of high activity driven and controlled by a computer system is stepping within hollow catheters which have been previously implanted into the tumor according to pre designed chart specific dwell time for a specific dwell position within a specific catheter The treatment time for such a temporarv implantation is very short Due to the high activity of the used single stepping source Iridium 192 nuclide this method is called HDR We consider here only the HDR case Clinicians need to know before starting inserting catheters needles into the tumor how many catheters are needed and what is their ideal position into the tumor The step described above for defining the appropriate geometry number and location of needles within the tumor in order to achieve the desired therapeutic effect is called pre planning We have a post implantation dose optimization problem or post planning problem i e given the position of the source dwell positions to determine the dwell times such that the resulting dose distribution satisfies to the maximum possible extent the objectives of conformal HDR brachytherapy Conformal anatomy based dose optimization considers the dose in the planning target volume PTV and critical structures including the surrounding nor
55. hra exceeds the critical dose The coverage of the PTV is 92 If we reduce it to 88 only 5 of the urethra receives a dose larger than Da This is the price we have to pay Some histograms of all solutions and the marked two solutions are shown We accept solution Nr 31 and calculate the isodose distribution WinOpt HDR User Manual 92 E Eononan 082 Ei Ba 0 000000 Ez Ad 035000 Er IE 22735000 0 823462 91 540000 000 12 360000 0 821097 91 305000 am 51 960000 0814961 90 860000 ooo00oo 20 290000 0810073 90690000 Mana 18 965000 ii git 0 509605 30 505000 0 000000 30 555000 MW IM 0 805059 90 255000 oop0000 24 595000 0 804722 90 390000 an 52 205000 0 799073 89 570000 goon 17 465000 orgs 89795000 MO 24 510000 0 785330 Bas mag 16 225000 Or BR 270000 mon 17 180000 1 779530 58 255000 0 DO E 900000 O F781 75 57 975000 0 000000 43 570000 0 778001 BB o ooooo0 13 810000 C0774516 87920000 ooooo00 11 720000 0 774415 Ee S65000 ana t 4 155000 D Dret WinOpt HDR User Manual 93 Histograms Display COIN All Histas 1h 00 Display Diff DVH urethra AEE gg peseur FS were poo fow Save Image x i Save Image
56. ich Switzerland edited by E Zitzler K Deb L Thiele C A Coello Coello D Corne Lecture Notes in Computer Science Vol 1993 Springer 167 180 2001 Abstract WinOpt HDR User Manual 99 M Lahanas N Mihckovis M Papagiannoponlon K Karonzakis D Baltas and N Zamboglon Application of a Hybrid NSGA II Multiobjective Algorithm for Anatomy based Dose Optimization in Brachytherapy EUROGEN 2001 Evolutionary Methods for Design Optimisation and Control with Applications to Industrial Problems Athens Greece 19 21 September 2001 Abstract Preprint in pdf format N Mitckovic M Lahanas M Papagiannopoulou K Karouzakis D Baltas and N Zamboglon Application of a Multiobjective Genetic Algorithms in Anatomy based Dose Optimization in Brachytherapy and its Comparison with Deterministic Algorithms EUROGEN 2001 Evolutionary Methods for Design Optimisation and Control with Applications to Industrial Problems Athens Greece 19 21 September 2001 Abstract Preprint in pdf format WinOpt HDR User Manual 100 APPENDIX 1 Template Coordinate System Templates are accessories being used in brachytherapy for fixing helping positioning and guiding of catheters needles that have to be inserted into the tumour volume in the human body Following figure shows the coordinate system for templates in WinOpt The definition of the Template coordinate system is shown in the following two Figures Template Coordinate System
57. imization methods are single weighted objective methods the dose optimization problem is a true multiobjective problem and therefore multiobjective optimization methods should be used The gradient based algorithm due to its efficiency allows the construction of the so called Pareto or trade off surface which contains all the information of the competition between the objectives which is necessary for the planner to select the solution which best fulfills his requirements One problem of this algorithm is that the weighted sum as used in all conventional dose optimization algorithms cannot provide solutions in possible non convex parts of the Pareto tradeoff surface because a convex weighted sum of objectives converges only to the convex parts of the Pareto front Another major limitation of the algorithm is its restriction to convex objective functions for which gradients can be calculated In this case according to the Kuhn Tucker theorems a global optimum can be obtained and the entire Pareto front is accessible from the weighted sum If we search for an optimal set of importance factors dividing each importance factors in n points then the number of combinations for amp objectives is approximately proportional to and the shape of the entire trade off surface require a very large computational time Most realistic problems require the simultaneous optimization of many objectives It is unlikely that all objectives are optimal for a single set of par
58. le time if the number of different set of importance factors is not very small WinOpt HDR offers a true multiobjective optimization with evolutionary algorithms for variance and DVH based objectives Inverse planning tries to find an optimal distribution of source dwell positions such that using this distribution with a small as possible number of source dwell positions we can achieve a dose distribution which fulfills the planners optimization criteria as much as possible The actual distribution after the implantation may differ and or there may be additional a significant modification of the PTV and OAR geometry A goal of post implant optimization is therefore still to obtain a good solution WinOpt HDR User Manual 34 The User Interface A SD Taste e T Pima R ls ch Seleka ES tos ae j Tobbi mE Wieoye HOR hea Cat TH Flo 0 SS 4 F User Interface of WinOpt HDR 2 0 The main toolbar selects the module and the dialog on the right side The secondary toolbar is used for the bi objective Window and the Anatomy Window that shows the anatomy i e the VOIS catheters and sources and isodose distributions ea es eC A The Anatomy Window displays the VOIS catheters and the source distribution B The Main Toolbar selects the module and the corresponding dialog field C The Secondary Toolbar is used for functions of the Anatomy Window D The Dialog field is used to set paramet
59. m Check if points are double or contour polygons are not simple Check if there is any contact or intersection between the VOIS Collision Detection Test between PTV and BODY triangles Number of triangles in Object Nr 1 2254 Number of triangles in Object Nr 2 583 All contacts between overlapping triangles Number of box intersection tests 179 Number of contact pairs 0 Collision Detection Test between Rectum and PTV triangles Number of triangles in Object Nr 1 2254 Number of triangles in Object Nr 2 2124 All contacts between overlapping triangles Number of box intersection tests 13 Number of contact pairs 0 Here we have a contact between the urethra and PTV Collision Detection Test between urethra and PTV triangles Number of triangles in Object Nr 1 2254 Number of triangles in Object Nr 2 528 WinOpt HDR User Manual All contacts between overlapping triangles Number of box intersection tests 1631 Number of contact pairs 85 contact O tri 261 and tri 39 contact 1 tri 261 and tri 38 contact 2 tri 261 and tri 37 contact 3 tri 261 and tri 25 contact 4 tri 260 and tri 39 contact 5 tri 260 and tri 41 contact 6 tri 260 and tri 40 contact 7 tri 260 and tri 25 contact 8 tri 260 andtri 24 lt lt List of all triangles in contact contact 9 tri 258 and tri 41 contact 10 tri 259 and tri 41 contact 81 tri 2201 and tri 520 contact 82 tri 2201 and tri 508 contact 83 tri 2201 and tri 509 contact 84 t
60. mal tissue Goals of the conformal anatomy based dose optimization is the complete coverage of the PTV with a dose at least equal to the prescription dose and simultaneously to avoid dose values above some critical value specific for each critical structure in the surrounding normal tissue and in critical structures inside or in the neighborhood of the target An additional objective is the avoidance of dose values much above the prescription dose in the target With anatomy based optimization we also mean that user can define his target for the tumor and the dosimetric limitations to the critical structures For example for the case of prostate this could be deliver at least 7 0 Gy to the prostate gland where doses of 10 5 Gy have to be avoided to the urethra and dose of 10 Gy have to be avoided to the rectum Since these objectives are usually controversial the optimization system has to offer following tools User can define his own pre known significance factors of the objectives User can define several combinations of these significance factors 3 The system is able to estimate the feasible spectrum of solutions among which the user has to make his decision WinOpt HDR User Manual 9 Tools 1 and 2 have to be fast enough so that the results can be evaluated and used for the interactive steps Step 3 should be utilized for finally optimizing the implant dosimetry BODY Sampling Point i rner ace nn iaa i Catheter Dwell j
61. mple of all non dominated solutions Two main approaches can be used A weighted sum where each objective is combined with a weight A convex combination is formed and for each such combination a single objective optimization algorithm is used Each such point produces one possible non dominated solution To obtain a representative sample of solutions the optimization algorithm has to be applied several times As the number of objective increases the combinatorial possibilities increases exponential Since a convex sum is formed only convex parts of the Pareto surface are accessible by this method 3 The repeated application of the optimization requires a considerable time to obtain a representative sample of the Pareto front The other approach is a true multiobjective optimization method which does not use weights but a population of solutions is guided in the multidimensional space towards the Pareto front In order to prevent the population to converge to a small part of the Pareto surface evolutionary principles are used to distribute as uniformly as possible the population over the entire trade off surface WinOpt HDR User Manual 27 Solutions obtained with the use of weighted sums in optimization procedures The solution which is obtained in the conventional weighted sum approach depends on the shape of the Pareto front and the importance factors used In Fig 10 the set F is shown for two objectives f and f The importance factors of the w
62. ms DVH The differential DVH of a PTV shows the part of the volume with a specific dose value The optimal homogeneous dose distribution marked by a red line is a delta function and practical impossible to realize PTV DVH of optimal dose distribution 80 co D 60 DVH of real dose distribution mm gt a a eee eee wen 38 i o of Volume with D D gt 1 5 20 0 0 1 D ID 2 3 4 Fig 9 A cumulative DVH shows the part of a volume with a dose larger than a specified dose value The ideal cumulative DVH for the PTV is 100 up to the reference dose value and 0 above Objectives 1 2 and 3 are used such that the resulting dose distribution approaches as close as possible the optimal dose distribution The objective function 4 tries to satisfy the constraint for the dose inside critical structures WinOpt HDR User Manual 23 Single objective weighted sum optimization We support the conventional weighted sum approach of a single solution formed by the linear weighted sum of the individual objectives The weight of each objective is called also importance factor For two objectives f x and f x we have f wf x wha w and w are the importance factors of f x and f x respectively An objective will contribute more to the objective function f if it is associated with a large cotresponding importance factor guiding probably thus the search engine to a smaller value for this objective This is of course not always possibl
63. n These criteria are expressed as computable functions f x f x of the decision variables which are called objective functions These form a vector function f In general some of these will be in conflict with others and some will have to be minimized while others are maximized The multiobjective optimization problem can be now defined as follows Find the vector x X X Xy which will satisfy the m inequalities o x gt 01 1 2 the p equality constraints h x 0 1 1 2 P and optimize the vector function f The constraints define the feasible region X and any point x in X defines a feasible solution The vector function f x is a function that maps the set X in the set F that represents all possible values of the objective functions Normally we never have a situation like this in which all the f x values have a minimum in X at a common point x We have to establish certain criteria to determine what would be considered a optimal solution One interpretation of the term optimum in multiobjective optimization is the Pareto Optimum For two vectors x X X Xy and y V1 Vo Vn Of the same dimension eguahty and less than and greater than relationships are fulfilled if the relationships are true element by element A fourth partial less than relationship can be defined as follows x is partially less than y if Vi 1 N x lt y Ad te 1 N x lt y For minimization problems if x is partially less th
64. n e g 3D volume 2D on CT plane Method Begin Type Isotropic or Ellipsoid or Margins Begin X XXX XX X XXX XX Y XXX XX WinOpt HDR User Manual XXX XX Z XXX XX XXX XX End End End Connections Begin 106 Here enter information for connection amp connection type between two 123 25 23 12 12 34 Vector 1 123 25 23 12 12 34 Vector 2 123 25 23 12 12 34 Vector 3 123 25 23 12 12 34 End Volume 1234567890 123 Surface 1234567890 123 Inner Surface 1234567890 123 Number of Contours NCNVOI 1 Contour 0 Begin Type XXXXXX e g not on an Image or Number of Points NPCONVOI Point O Begin Coordinates 123 25 23 12 12 34 Image 1 End Point NPCONVOI 1 Begin Coordinates 123 25 23 12 12 34 Image 34 End End Contour NCNVOI 1 Begin Type XXXXXX e g not on an Image or 222 222 WinOpt HDR User Manual End Number of Points NPCNCNVOI 1 Point O Begin Coordinates 123 25 23 12 12 34 Image 1 End Point NPCNCNVOI 1 1 Begin Coordinates 4123 25 23 12 12 34 Image 34 End Number of Inner Contours NICNVOI 1 Contour 0 Begin End Type XXXXXX e g not on an Image or Number of Points NPICONVOI 1 Point O Begin Coordinates 123 25 23 12 12 34 Image 1 End Point NPICONVOI 1 1 Begin Coordinates 4123 25 23 12 12 34 Image 34 End Contour NICNVOI 1 Begin Type XXXXXX e g not on an Image or Number of Points
65. n Test between rectum and PTV triangles Number of triangles in Object Nr 1 536 Number of triangles in Object Nr 2 190 All contacts between overlapping triangles Number of box intersection tests 5343 Number of contact pairs 130 contact contact contact contact contact contact Oe ee ee tri tri tri tri tri tri 383 and tr 117 384 and tr 116 383 and tri 116 476 and tri 147 476 and tri 177 384 and tri 115 contact 127 tri 451 and tri 159 contact 128 tri 45 and tri 186 contact 129 tri 359 and tri 159 WinOpt HDR User Manual 96 General References 1 D Baltas C Kolotas K Geramani R F Mould G Ioannidis M Keckidi and N Zamboglou A Conformal Index COIN to evaluate implant quality and dose specifications in brachytherapy Int J Radiation Oncology Biol Phys Vol 40 No 2 512 524 1998 2 Baltas D Milickovic N Giannoul S Lahanas M Kolotas C Zamboglou N New Tools of Three Dimensional Imaging Based Brachytherapy in the Frontiers of Radiation Therapy and Oncology Series 59 70 2000 Abstract 3 I Das and J Dennis A Closer Look at Drawbacks of Minimizing Weighted Sums of Objectives for Pareto Set Generation in Multicriteria Optimization Problem Structural Optimization 14 No 1 1997 4 G K Edmundson Geometry based optimization for stepping source implants in Brachytherapy HDR and LDR A A Martinez C G Orton and R F Mould eds
66. n of source dwell positions This is called post implant optimization and is an answer to the problem b WinOpt HDR provides a unique methodology in that it does not necessary needs arbitrary importance factors for the various objectives but provides the entire spectrum of possible solutions for the various competing objectives Also it does not produce negative and non physical dwell times It is surprising that the majorities of treatment planning systems produce such infeasible solutions and then apply a simple correction mechanism which deteriorates the quality of the solution Most of the treatment planning systems use arbitrary rules which although have been gained through experience are not able to give always the best result The methods in WinOpt HDR although expect from the planner more interaction in order to obtain the best result for the patient This interaction is mainly in the selection of the best solution out of the many competing solutions WinOpt HDR used DVH based objectives for the inverse planning optimization WinOpt HDR User Manual 33 Post Implant Optimization Post Implant Optimization assumes that the source dwell positions are given WinOpt HDR supports deterministic single and multiobjective optimization with variance based objectives Additional DVH based aggregate single objective optimization is supported with fast simulated annealing Multiobjective optimization can in principle also be used but requires considerab
67. nsenenenenennenennn 94 GOCMCTOL AC CLE CHICES ex een er 96 WinOpt HDR User Manual 4 WiNOBEHDR References 98 APPENDIX 2uennuaearririr 100 1 Template Coordinate System aan isinai 100 2 The VOI Charisma Files ue0uaseenbaansn ars 101 3 Ihe Catheters Charisma File ae AEREE 109 4 The Loading Charisma Files a 112 The WINOPEHDR 108 Pille na a RING RE GEHE 115 The WinOpt HDR Solutions File u uus nuckeek a 120 WinOpt HDR User Manual 5 WinOpt HDR Overview and Tutorial WinOpt HDR is a software toolkit designed for anatomy based dose optimization for high dose rate HDR brachytherapy Its algorithm are continuously updated and improved It has been developed by Michael Lahanas in cooperation with Dimos Baltas head of the Medical Physics amp Engineering department of Klinikum Offenbach in Germany Kostas Karouzakis and Stavroula Giannouli from the National Technical University of Athens helped in the development of parts of the WinOpt HDR source code Additional studies by Natasa Milickovic and Maria Papagiannopoulou were important in estimating the performance of various algorithms Some of the methods which are unique in HDR brachytherapy have been described in numerous publications in medical physics journals and in conference proceedings WinOpt HDR is a demonstration toolkit for our solution which we offer for the true multiobjective dose optimization problem in HDR brachytherapy It does not include other tools such as
68. of cumulative and differential dose volume histograms obtained from the sampling points used in the optimization and for comparison the corresponding histograms with a large number 20000 of sampling points Here is a example WinOpt HDR v 2 00 Anatomy based HDR 3D Dose Optimization WinOpt HDR by Michael Lahanas Email mlahanas gmx de www www mlahanas de Developed by Michael Lahanas Kostas Karouzakis Stavroula Giannouli Maria Papagiannopoulou and Natasa Milickovic from Medical Physics and Engineering Department in Offenbach Germany under the supervision help and support from Prof Dimos Baltas Monte Carlo Dosimetric Look up Table from Pantelis Karaiskos et al University of Athens OS time 12 02 17 OS date 10 22 01 Monday 22 October 2001 Dose Statistics Template Based Optimization ACTIVE NEEDLE MAP 8 00 2 7 50 3 gt gt 2 gt 3 7 00 2 2 6 50 z gt 3 6 00 2 j 5 50 5 00 s 3 2 4 50 z X X X 2 4 00 3 50 2 s X X 3 00 s 2 50 X s X 2 00 gt 5 1 50 1 00 i X XxX X gt lt I I x WinOpt HDR User Manual 121 ACTIVE DWELL MAP A a B b C c D d E e F 8 00 0 0 0 0 0 0 0 0 0 0 0 7 50 0 0 0 0 0 0 0 0 0 0 0 7 00 0 0 0 0 0 0 0 0 0 0 0 6 50 0 0 0 0 0 0 0 0 0 0 0 6 00 0 0 0 0 0 0 0 0 0 0 0 5 50 0 0 0 0 0 0 0 0 0 0 0 5 00 0
69. ooooooo0oO00o060o0 lt Ooooooooooo0o0o0o00o0 This table shows the number of source dwell positions on each catheter on the template grid that can be used or are selected WinOpt HDR User Manual 47 FREE LENGTH cm A a B b C c D d E e F 1 out out out out out out out out out out out 1 5 out out out out out out out out out out out 2 out out out out out 10 30 10 30 out out out out 2 5 out out 11 58 10 30 10 30 10 30 10 30 10 30 10 30 out out 3 out out out 10 30 10 30 10 30 10 30 10 30 10 30 out out 3 5 out out 10 30 10 30 10 30 10 30 10 30 10 50 10 32 out out 4 out out 10 30 10 30 10 30 10 30 10 30 out out out out 4 5 out out 10 30 10 30 10 30 10 30 out out 10 61 10 33 out 5 out out 10 30 10 38 10 30 10 30 10 30 10 30 10 30 10 30 out 5 5 out out 11 55 10 30 10 30 10 30 10 30 10 30 10 30 10 30 out 6 out out 11 55 10 30 10 30 10 30 10 30 10 30 10 30 10 30 out 6 5 out out out 10 30 10 30 10 30 10 30 10 30 10 30 10 30 out T out out out out 10 30 10 30 10 30 10 30 out out out 7 5 out out out out out out out out out out out 8 out out out out out out out out out out out This table shows the free length of each catheter on the corresponding template grid point and determines therefore how deep each catheter should be inserted WinOpt HDR User Manual 48 Template View The template view is used to display the VOls and the catheter position on the template plane Source dwell position VOI C atheter Y Template Global co
70. ordinate system y The template view The contours of VOI are shown along the normal to the template plane up to the maximum needle depth z min WinOpt HDR User Manual 49 WinOpt HDR offers a manual method for the selection of catheters to be implanted based on the VOls and template geometry It offers an auto activation mechanism that ensures that no source dwell position will be closer than a given distance to a specified VOI surface The first step finds the catheters that can be used considering OARs in the path of each catheter Usually the number of catheters which pass through the PTV can be very large If the template grid is very fine catheters on a larger grid using each second grid point can be used Additional each possible catheter can be included manually Tem plate WIEN Active 7 Show 77 0900 i PTV j Fe a FR m Template i 5 5 Say Loading 5 0 5 0 Catheter density 4 5 4 5 EHEN 4 0 4 0 ane nae Catheter Mr 3 0 a A80 26 2 9 Source Nr 2 0 2 0 LE 1 5 1 0 1 0 Dialog which shows the grid of the template the catheters and VOls at various distances from the template The selected catheters are shown in red The catheters which can be selected in light blue while the catheters which are shown in dark gray at the give position can not reach the selected plane WinOpt HDR User Manual 50 SE a ee oe m Displays the template Displays the used catheter on the template grid
71. ped a method to uniformly distribute sampling points on the whole PTV surface This method requires the triangulation of the entire PTV surface Based on the stochastic universal sampling method uniform distributed sampling points are generated on the PTV surface see 9 In the past treatment planning systems using dose point based optimization methods had to consider the problem of negative dwell times Many phenomenological approaches have been used to eliminate reduce or correct negative dwell times either at each optimization step or at the end of the optimization One method specifies constraints such as the difference of dwell times between closely situated sources 4 14 Therefore the objective function is extended including a term that considers the gradient of dwell times differences between neighbor sources Since all algorithms use importance factors some empirical importance factor has to be used for this additional objective This approach includes the additional objective like a penalty function that tries to reduce the number and the magnitude of negative dwell times It does not eliminate such solutions and finally always a correction method is necessary This method just sets all negative dwell times equal to 0 and renormalizes the dwell times after this correction This method has the following problems It includes an additional objective function that tries to reduce or eliminate negative source dwell times This approach reduces
72. point is on the PTV surface If the dose is normalized using the PTV dose average value then we cannot have always high enough dose values anywhere in the PTV As long as there are no sources outside the PTV it will more or less help us to avoid high dose values in the surrounding tissue Also we have not full control of the dose in the organs at risk This method assumes that the reference isodose can actually have the shape of the PTV As the shape of the PTV is usually complicated and isodose surfaces are less complex it requires that the number of sources is large and adequately distributed in the PTV A method that is less restricted is the normalization free method In this method there is in principle no point defined at which the dose is normalized The optimization is free so that finally we have the best possible results Currently there are some requirements to document the dose in a definite way and the established methods introduce constraints that reduce the quality of the dose distribution which can be obtained It is still not clear that the problem is not a single objective problem but a multiobjective problem with contradictory and competing objectives that requires the determination of at least a portion of the Pareto front Today when there are no exact established quality criteria the planner cannot guide the system exactly to a single solution on the Pareto front The problem is transformed into a single objective problem using a
73. r is shown the population of solutions In green the accumulated population of non dominated solutions In red the population members which are selected in the Filter Dialog Save image of Pareto window as a TIFF file Display distribution of archived solutions in Pareto window Display of all solutions Size of solution in the Pareto window Ext Size 60 Reduce external archived population to a given size Uses a hierarchical cluster reduction technique WinOpt HDR User Manual Template based inverse Plan Optimization Template based Multiobective Evolutionary Inverse Plan Objectives for the organs at risk 71 WinOpt HDR User Manual 72 The Dose distribution Dialog Eu g Proa Er 00 a Tia 500 om 1 2 2 000 J Cy 3000 21 am 13 S00 ae LT ta 000 ll Cl E Cr Eoo en oo AE SO St ol This dialog is used to select the two and three dimensional isodoses to be displayed with the VOIS in the Anatomy Window Switches on off all the checked 3D isodoses if an isodose calculation has been completed Switches to a solid rendering of the 3D isodose surfaces blue cyan green magenta fed ammen pa m ag nn mn mn Set color mapping for 2D and 3D isodose lines and surfaces Show in Anatomy Window color mapping spectrum WinOpt HDR User Manual 5 000 4 500 4 000 3 500 3 000 2 500 2 000 1 500 1 000 0 500 Min 2000 Limits Set Min and Maximum i
74. r of objectives is now five it is recommended to increase the number of individuals so that we have a representative set of the Pareto surface We could impose a constraint on the PTV coverage PI V that would select solutions that satisfy this constraint Here we want to see the entire Pareto surface If we set PTV lt 0 3 then with a high probability we accept only non dominated solutions for which the coverage of the PTV with the prescription dose is larger than 70 After the optimization run we use the filter dialog to select a solution from the non dominated set We set constraints for the solutions for the urethra and the rectum we allow a maximum of 3 of the volume to exceed the critical dose Actually if we set both constraints equal to 0 we see that the maximum the possible coverage of the PTV is 74 For the conformity index we set the constraint that it should be larger than 0 85 We select solutions with coverage of the PTV larger than 85 We obtain 11 non dominated solutions WinOpt HDR User Manual 0 209330 023754 0 25249 0 25083 0 29568 0 27243 0 31728 0 30393 0 28405 0 32226 032558 0 32509 0 28614 0 27503 0 27918 0 26822 0 27151 0 26040 0 27730 0 26104 0 24701 0 25075 0 00000 0 00000 0 0000 0 00000 0 00000 0 00000 0 00000 0 000 2 00000 0 00000 0 00000 0 02786 0 00857 0 00000 0 03143 Ort 0 00000 0 01429 0 0 00000 0 00286 0 00000 84 1 000000
75. rbitrary importance factors for each objective and combining the objectives into a single objective function to be minimized What does it mean to use an importance factor 0 9 for objective 1 and 0 1 for objective 2 That the objective 1 is more or less 9 times more important than objective 2 If there is a strong tradeoff then a combination of 0 89 and 0 11 can give quite different results Some algorithms require importance factors of 20000 for one objective and 10 for another which shows that also the optimal importance factors depend on the objective function used In IMRT we have found that the surrounding tissue and PTV coverage objective produce not the best results with a 0 1 and 0 9 importance factor respectively as assumed but sometimes the best result was obtained with importance factors 0 7 and 0 3 or 0 3 and 0 7 respectively This is because some objectives are cooperating to some extent and to some extent they are competing depending on the importance factors of the other objectives and in a complicated way which depends on the WinOpt HDR User Manual 11 topology of the PTV and the OARs the beam orientation etc It is not very much different in brachytherapy Not only the optimization method is important but also the distribution of sampling points used in the optimization procedure One problem is that dose variances are calculated from sampling points limited on the contours which even if their number is large it is not true that the va
76. res can be considered The critical dose value in fractions of the reference dose value must be supplied Apply a PTV variance based dose optimization using variances Apply a DVH based dose optimization CS objectives are included only as constraints Only a single constraint is supported All the CSs are included in one single constraint WinOpt HDR User Manual Settings for the DVH based multiobjective dose optimization Parameter Display Temporary Generate opimo fobabiity H1 nder Parameters for mutation Population Size lt 300 fo Specify population size 100 Generations Specify number of generations 10 Initialization of pe Members Specify number of solutions to be initialized by a deterministic algorithm 55 WinOpt HDR User Manual 56 The Simulated Annealing Optimization Dialog Simulated Annealing Parameter F a This dialog is for the multiobjective weighted sum optimization method based on the fast simulated annealing algorithm Two methods are supported PTV based and DVHs based For both methods either a single optimization run can be used i e a set of importance factors for each objective must be supplied For multiobjective optimization runs using the weighted sum it is possible to generate many solutions from successive runs The importance factors of each run are calculated from the number of intervals parameter set by the user This parameter defin
77. ri 2201 and tri 510 Contours of all VOIS imported Organ BODY Slice 0 Perimeter 367 895706 Points in this slice 9 anti CCW order 59 6700 110 3100 0 0000 59 6700 10 2100 0 0000 139 1700 10 2100 0 0000 138 9700 110 3100 0 0000 102 4600 60 8500 104 0000 Active sources and their position in the catheters Number of Active Sources found 294 Act Dwell Nr 1 Catheter Nr 1 Dwell Nr 1 Pos 89 900166 60 218575 94 800000 Act Dwell Nr 2 Catheter Nr 1 Dwell Nr 3 Pos 89 900166 60 218575 84 800000 Act Dwell Nr 3 Catheter Nr 1 Dwell Nr4 Pos 89 900166 60 218575 79 800000 Act Dwell Nr 293 Catheter Nr 46 Dwell Nr 1 Pos 114 770972 90 424905 78 550000 Act Dwell Nr 294 Catheter Nr 47 Dwell Nr1 Pos 99 750094 95 300897 74 800000 List of all catheters and their geometry defined by catheter describing points Number of Catheters 47 Catheter 1 List of Describing points Number of Describing points for this catheter 2 Point 1 Pos 89 9002 60 2186 9 8000 Point 2 Pos 89 9002 60 2186 100 8000 List of dwells Number of dwells in this catheter 17 Act Dwell 1 Pos 89 9002 60 2186 94 8000 t 0 00000 0 00000 1 00000 Inac Dwell 2 Pos 89 9002 60 2186 89 8000 t 0 00000 0 00000 1 00000 Act Dwell 3 Pos 89 9002 60 2186 84 8000 t 0 00000 0 00000 1 00000 Act Dwell 4 Pos 89 9002 60 2186 79 8000 t 0 00000 0 00000 1 00000 Act Dwell 5 Pos 89 9002 60 2186 74 8000 t 0 00000 0 00000 1 00000 Inac Dwell 6 Pos 89 9002 60 2186 69 8000 t 0 0000
78. riances are approximated with sufficient accuracy Another problem is the use of points inside catheters Optimization is the most important part of a treatment planning system and to a smaller extent for example its volume rendering capabilities It seems therefore strange that these constraints used in the optimization models based on crude approximations are still used in the majority of planning systems It is true that speed is an important factor Nevertheless also methods such a geometric optimization although may be fast require from the planners often to manually intervene in order to modify the dwell times if the results are not satisfactorily Using only a very small set of importance factors 1 2 usually the planner is left without information of what actually is possible The true multiobjective optimization is free of importance factors Even if a set of importance factors which somehow gives reasonable results can be found for some fixed topology it requires time to find these and a significant time has to be spent A multiobjective algorithm does not require such training Sometimes although it seems strange in principle it is easier to calculate all possible solutions at once than only one particular This is true for true multiobjective optimization algorithms In a single objective optimization there is only a single decision accept the solution or not If not a new solution must be calculated or a manual manipulation of dwell time
79. roduced very close to the source dwell positions Statistical values obtained from the sampling points are calculated therefore with a higher accuracy 9 Catheter describing point Outer radius R M of catheter Margin A Fig 4 A catheter is defined by catheter describing points These points are connected with cylinders and at each catheter describing point an additional sphere is included The set of catheters cylinders and spheres are used to describe the geometry of a catheter that may be either metallic linear or plastic and curved WinOpt HDR User Manual 15 Extended bounding box Bounding box E BE F E E E E a E E E E F Catheter radius R Fig 5 The set of all catheters defined by the catheter describing points and extended by the catheter radius defines the bounding box of all catheters The orientation of this bounding box is optimized such that its volume is minimum Only sampling points in this optimal oriented box are tested if they are inside the catheters WinOpt HDR excludes the volume of parts of catheters and of parts of OARs which are inside the PTV Such a case we have for example in a prostate implant where the volume of the urethra and the catheters inside the prostate is ignored and not included in the prostate volume WinOpt HDR User Manual 16 Optimized bounding boxes Our anatomy based optimization method uses optimal oriented bounding boxes 10 for a variety of rea
80. rom sampling points used in optimization Dose D_ref counts 0 050000 0 150000 0 250000 0 350000 0 450000 0 550000 0 650000 0 750000 0 850000 0 950000 1 050000 1 150000 1 250000 1 350000 1 450000 1 550000 1 650000 9 450000 9 550000 9 650000 9 750000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 1 054439 13 707713 22 143229 27 415427 40 068700 51 667535 49 558656 56 939732 55 885293 50 613095 50 613095 0 000000 0 000000 0 000000 0 000000 WinOpt HDR User Manual 9 850000 0 000000 9 950000 0 000000 CDVH from sampling points used in optimization Dose D_ref percent 0 000000 0 100000 0 200000 0 300000 0 400000 0 500000 0 600000 0 700000 0 800000 0 900000 1 000000 1 100000 1 200000 1 300000 1 400000 1 500000 1 600000 1 700000 1 800000 1 900000 8 900000 9 000000 9 100000 9 200000 9 300000 9 400000 9 500000 9 600000 9 700000 9 800000 9 900000 100 000000 100 000000 100 000000 100 000000 100 000000 100 000000 100 000000 99 800000 97 200000 93 000000 87 800000 80 200000 70 400000 61 000000 50 200000 39 600000 30 000000 20 400000 15 800000 13 600000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000
81. s The tasks should perform in the following order Preparation and input of data for the optimization Selection of a optimization method Selection of a solution Analysis of the results a A It is assumed that the final acquired anatomy definition of PTV and critical structures contours with the needles in situ CT or 3D U S reconstructed 17 18 Additional the source dwell positions to be considered in the optimization are selected Thereafter the anatomy based optimization tool as described previously can be utilized for the case of HDR brachytherapy for defining the exact source dwell times and source dwell positions within the implanted catheters that will result to the desired 3D dose distribution user objectives The parameters needed to drive the afterloader in order to deliver the planned brachytherapy are then calculated and stored WinOpt HDR User Manual 6 It is assumed that the geometry of the VOls the catheter geometry and the source dwell positions and the dwell times or loading is written in a special format see Appendix 2 and 3 Minimum System Requirements WinOpt HDR version 2 0 runs on a standard PC hardware platform with the operating system Windows 95 98 Windows NT or Windows 2000 Dosimetric fast calculation requires the use of large look up tables ARAM of 256 MB and more is therefore recommended A graphic card that supports OPENGL is recommended but not necessary Current version supports onl
82. s or dose rescaling remains which takes a lot of time usually much more than the automatic optimization These are definitive not methods of the 21 century Today the computing power of modern PCs allows the determination of hundreds of possible solutions The planner now has information that was to him her not available before or only to a very limited extent The optimization problem is not an optimization problem anymore that has to consider arbitrary importance factors arbitrary correction methods and objectives based on nomograms or other hand waving arguments but is now transformed to a decision making problem The planner is confronted with all possible solutions and information He She knows now all the answers to the questions What is possible if I modify the importance factors What coverage can I get How this affects the dose value in the organs at risk Even if some new systems give an approximate answer to some of these questions these methods use a sometimes complicated way to obtain part of this information A true multiobjective method requires a set of objectives functions that are more intuitive for the planner than variances of dose distributions The most natural way from a dosimetric based approach seems to be the use of dose volume histogram based objectives Methods have been proposed which try to optimize the importance factors using an optimization method The objective function for a set of importance factors is calcula
83. sodose values Press limit to set actual values Update isodose graph for values set in 5 000 4 500 4 000 3 500 3 000 2 500 2 000 1 500 1 000 0 500 Win Opt Example of 2D isodose lines 73 WinOpt HDR User Manual 000 4 500 4 000 2 000 1 500 1 000 0 400 2 in Opt Example of 3D isodose surfaces 000 4 500 4 000 2 000 1 500 7 000 0 400 Cc Wind pt Example of 3D isodose surfaces in solid mode and transparent 74 WinOpt HDR User Manual 75 The Volumetry Dialog This Dialog displays information about the volume and surface of the PTV body and organs at risk WinOpt HDR User Manual 76 The 3D View Dialog BD View UN fs fam EL _ Contour Points Box Orig Box Opt Box STR JE Dwelson Swale tr LJ Catheters Dwell Tana Template mase a JE Line Width 1 RESET View More Setti The 3D View dialog controls the display parameters of the Anatomy Window VOIS bounding boxes Catheters sources and sampling points can be switched on or off Display original not oriented bounding box of VOI Box Opt Display of optimal oriented bounding box of VOI Display strata if a stratified sampling method is used Display of activated source dwells positions Display catheters Display of not activated source dwell positions Display tangents at dwells which shows the direction of the cylindrical so
84. sons For the efficient calculation of dose volume histograms DVHs of anatomical structures obtained from sampling points inside the three dimensional triangulated objects using the test routines described by Lahanas et al 9 The sampling points are required to be outside catheters and critical structures but inside the object The generation time of sampling points can be reduced significantly using optimal oriented bounding boxes of the object the critical structures and catheters For an optimal orientation of dose calculation grids for fast Fourier transform FFT based dose calculation methods 8 the number of grid points inside an object have to be maximized If the axes of the grid are given by the axes of the minimum bounding box of the object then the number of grid points inside an object is maximal and this significantly increases the efficiency and accuracy of this method For the efficient determination of the volumes of objects that take into account catheters a simple hit or miss Monte Carlo method with a large number of random points gt 10 inside the bounding box of each structure e g organs is produced The ratio of the number of points inside the structure to the number of points inside the associated bounding box which is maximal for minimum bounding boxes is equal to the ratio of the structure volume to the bounding box volume From the known bounding box volume the volume of the structure is then calculated For
85. t s 0 740380 x 99 750 y 95 301 z 69 800 List of all catheters and the total dwell time in each catheter Catheter Nr 1 Total Dwell Time 0 747730 s Catheter Nr 2 Total Dwell Time 2 727544 s Catheter Nr 3 Total Dwell Time 1 322610 s Catheter Nr 14 Total Dwell Time 1 956138 s Catheter Nr 15 Total Dwell Time 0 872949 s Catheter Nr 16 Total Dwell Time 0 740690 s Total Time for Treatment 32 483522 s 122 WinOpt HDR User Manual 123 Here are the relative dwell times normalized to unity at the maximum dwell time Normalized to unity dwell position weights Nr Nr Nr Nr Nr Nr Nr Nr Nr Nr 1 Catheter 2 Catheter 3 Catheter 4 Catheter 5 Catheter 90 Catheter 91 Catheter 92 Catheter 93 Catheter 94 Catheter 1 Dwell 1 Dwell 1 Dwell 1 Dwell 2 Dwell 15 Dwell 15 Dwell 15 Dwell 16 Dwell 16 Dwell 1 Weight 0 179915 x 89 900 y 60 219 z 94 800 3 Weight 0 123916 x 89 900 y 60 219 z 84 800 4 Weight 0 066100 x 89 900 y 60 219 z 79 800 5 Weight 0 100997 x 89 900 y 60 219 z 74 800 1 Weight 0 310696 x 99 900 y 60 301 z 92 300 1 Weight 0 168005 x 114 771 y 90 425 z 78 550 2 Weight 0 000031 x 114 771 y 90 425 z 73 550 3 Weight 0 381756 x 114 771 y 90 425 z 68 550 1 Weight 0 000195 x 99 750 y 95 301 z 74 800 2 Weight 0 466298 x 99 750 y 95 301 z 69 800 The normalization is mean dose on t
86. tains information about the dwells which should be used and their corresponding dwell times FEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE EEE EEE EEE EET H Catheters Source Loading Data File CHARISMA R Vs 1 0 0 1 C opyrights Pi Medical Ltd All Rights protected LEER ul a si mia ac vll a vl hal u a u a nl a al m al a al a vl a vl a a EEE EEE EEE EET H Source Positions X Y Z Usage Status Weight Usage Status Inactive Active Weight Inactive 0 000000 Active X XXXXXX if no optimization done 1 000000 If optimization done then Weight is in 0 1 All coordinates according to the World DICOM Coordinate System All dimensions are in mm Date is given in mm dd yyyy Time is given in hh mm ss PEER EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE EEE EEE EEE EEE HHT CHARISMA Software Version XX XX XX XX Loading File Version XX XX XX XX File creation date and time mm dd yyyy hh mm ss KHHHHHHHHH HHH HHH H Patient Name SurName GivenName Patient ID XXXXXXXXXXXXXXX Number of Catheters N Source Step Begin Catheter 0 Begin XX XXX End Catheter 1 Begin XX XXX Catheter N 1 Begin XX XXX End End Source Positions Begin Catheter 0 Begin Number of Points WinOpt HDR User Manual 113 NSO Point O Begin Coordinates 145 15 234 38 223 00 Status Inactive Weight 0 000000 End Point 1 Begin Coordinates 145 15 234 38 223 00 Status Active Weight 0 123456 Point NSO 1 Begin Coordinates 144 60 234 46 218 00
87. ted and then a different set of functions combined with another set of new importance factors is used to establish a quality criterion of the optimization results Here the problem has been bypassed by the introduction of a new set of artificial importance factors but this method is probably better than using a fixed set of arbitrary importance factors for the objective functions WinOpt HDR User Manual 12 It is clear that methods such as the geometric optimization method have been established through the years The hope is that the new optimization methods slowly will be introduced in new planning systems and the planners will get familiar with the new methods It requires some training but the result will be less arbitrary without unnecessary correction methods less manual intervention It will also produce better optimization results Finally it will produce probably new quality criteria that at the moment are not very well established It is true that we cannot expect miracles from any optimization method since the number of sources and their distribution finally introduces physical restrictions of what is possible Aim of a true multiobjective anatomy dose optimization is to show these limitations independent on arbitrary corrections and importance factors It is also necessary not to use the word inverse planning for only post plan optimization since inverse planning means actually to find not only the dwell times of the optimal solution but also
88. the cylindrical source axis and in opposite direction to the source drive cable see Fig 2 Fig 2 Source dwell positions for a prostate implant The white points are active source dwell positions i e are considered in the optimization process while blue points show inactive source dwell positions rejected by the auto activation algorithm A tangent vector at each source dwell position is shown which defines the direction of the cylindrical source and which is used for the dose calculation WinOpt HDR User Manual 14 Sampling Points We estimate the dose distribution inside the PTV critical structures and the PTV surface from the dose of a small number of points sampling points Fig 3 Sampling points distributed on the contours and on the triangulated surface For the contour based method no points are on both ends of the PTV Therefore a large part of the surface is undefined for the optimization algorithm and the resulting isodose is bounded only by the PTV contours Sampling points in the volume are generated from low discrepancy sequences or quasi random distributed sampling points In contrast to pseudo random distributed sampling points voids and cluster are avoided Monte Carlo generated quantities convergence much mote rapidly than a conventional pseudo random sequence We exclude sampling points inside catheters This reduces the influence of very large dose values of sampling points that occasionally are p
89. the quality of the solutions since it introduces some constraint on the solutions such as the difference of source dwell times between two neighbor sources Additional it requires an importance factor which cannot be unique and the influence on the result of it is case dependent Another problem is that it finally cannot avoid negative source dwell times and therefore it has to correct the solution with the non physical negative dwell times by setting these equal to 0 This further reduces the quality of the obtained solution WinOpt HDR User Manual 19 Other approaches use a constraint of positive dwell times They do not require some rather arbitrary additional objective function This in principle increases the number of optimization parameters by a factor of two If the constraints are satisfied only partial and negative dwell times are present in the final solution again a correction mechanism must be applied In order to eliminate this problem we use a simple mapping technique that transforms the linear constrained problem into a quadratic non constrained problem With this approach negative solutions are avoided No constraints or bounds are applied in the search space Therefore the method does not introduce any bias to the search algorithm No additional objective functions constraints correction or importance factors are required If organs at risk close or in the target except the surrounding normal tissue don t ned to be considered t
90. unoglou Catheter autoreconstruction in computed tomography based brachytherapy treatment planning Med Phys 27 1047 1059 2000 17 N Milickovic D Baltas S Giannouli M Lahanas and N Zamboglou Algorithm for autoreconstruction of catheters in computer tomography based brachytherapy treatment planning IEEE Transactions on Biomedical Engineering 48 No 3 372 383 2001 18 H Szu and R Hartley Fast simulated annealing Phys Lett A 122 157 162 1987 19 A Tsalpatouros D Baltas et al CT based Software for 3 D Localization and Reconstruction in Stepping Source Brachytherapy IEEE Transactions in Information Technology in Biomedicine 1 No 4 229 242 1998 WinOpt HDR User Manual 98 WinOpt HDR References M Lahanas D Baltas and N Zamboglon Anatomy based three dimensional dose optimization in brachytherapy using multiobjective genetic algorithms Med Phys 26 1904 1918 1999 Abstract Preprint in pdf format M Lahanas D Baltas N Mihckovi S Giannouh and N Zamboglon Generation of uniformly distributed dose points for anatomy based three dimensional dose optimization in brachytherapy Med Phys 27 1034 1046 2000 Abstract Preprint in pdf format S Gtannouh D Baltas N Mihckovi M Lahanas C Kolotas N Zamboglon N Uzunogla Autoactivation of Source Dwell Positions for HDR Brachytherapy Treatment Planning Med Phys 27 2517 2520 2000 Abstract Preprint in pdf format
91. urce Displays the Template Display of sampling points on the triangulated PTV surface Select font for text display that uses the standard font dialog WinOpt HDR User Manual 77 EI AP Algenan EB Allegro BT m Amel ype Md BT TF Arial Black Anal Narrow WinOpt HDR User Manual 78 HERT x Show sampling points in corresponding VOI 4 Select color uses the standard windows color selection dialog Cer ime mmm LET E ma ia i i m h m TELL mm m gr mn _ Es Show the settings dialog for more settings of additional display parameters Papy Show corresponding VOI re Point Size Set points size Line Width Set line width 79 WinOpt HDR User Manual WinOpt HDR User Manual 80 Examples of using WinOpt HDR PTV based optimization method We consider here an example of the dose optimization of a prostate implant where no critical structures except the surrounding normal tissue have to be considered We use the PTV based algorithm Conventional We select a multiobjective optimization with 21 solutions i e the importance factors are varied in steps of 0 05 units from 0 to 1 2 Lal a ri z rt mn Po a a a a Deterministic Post Plan Dose Optimization t mj TL re z i E Aa iat fates ey all ej Al I a BE br rt We see the list of all solutions in the Filter dialog WinOpt HDR User Manual 81 CONTT OO0 DYHELOOc oRP TY D
92. ves towards the Pareto front and as it samples at the end non dominated solutions WinOpt HDR User Manual 65 The Analysis of solutions Dialog This dialog is used for the calculation of DVHs and statistical parameters of the dose distribution in the VOIS and on the PTV surface It is used also for the calculation of a selected solution In case of many solutions obtained from various optimization runs or from a multiobjective optimization algorithm solutions can be save and compared A decision tool is used to select a solution that satisfies the objectives of the planner Additional it helps the planner to understand the trade off between the objectives for the particular implant ccept the selected number This is necessary to confirm that the selected Nr in the following dialog element should indeed be selected e H Number of solution to be selected Displays the Filter dialog for the selection of solutions from a set of non dominated solutions Includes critical structures in the calculation of the COIN presented in the list of solutions WinOpt HDR User Manual 66 After selection and acceptance of this solution press this to generate output with statistics DVHs dwell times etc Additional it calculates the isodose surfaces WinOpt HDR User Manual 67 ns eee COIN D00 DVH 1 500 urethra D HI1 500 rectum 0 668835 10 655352 0 655276 0 642545 0 641206 0 640383 0 620393 0 619687 0 609008 0 599971
93. w we have to pay this by a reduction of the coverage of the PTV with the prescription dose It is now 90 9 WinOpt HDR User Manual 82 Dose distribution for Solution Nr 6 in the Filter Dialog The dose distribution shows that the dose variance in the volume has been reduced therefore also high dose values inside the PTV Of course the optimization does not consider critical structures but this method can be used to some extend also for critical structures 100 0 a0 0 60 0 40 0 20 0 1 2 3 IrDret 4 1 2 3 DeDret 4 Display Cum DYH of PTY il Color Display Cum DVH of Urethtra mea a rons Eil WT enalie lt fpo feo Snag v em foon H en po p save Image Abbrechen We finally look at the Histograms The solutions 0 and 6 are marked For solution 0 there are considerably high dose values in the PTV and in the urethra This is reduced significantly for solution Nr 6 WinOpt HDR User Manual 83 DVH based multiobjective optimization method We consider here an example of the dose optimization of the prostate implant used previously where we now consider two additional critical structures rectum and urethra We use the evolutionary true multiobjective optimization algorithm and use the DVH based objectives We set the critical dose values for the rectum and urethra as fractions of the prescription dose value We require 300 non dominated solutions As the numbe
94. wo methods supported a using uniform and b random distributed importance factors WinOpt HDR User Manual 62 Uniform Weights With uniform distributed importance factors or weights the Pareto set is sampled uniformly if it is convex and the algorithm is not trapped by local minima Uniformly distributed Weights are generated 11 beste Specify number of weights per objective Please consider that number of final solutions increases exponential with number of objectives Eq 7 and only maximal 500 solutions are currently supported Example of the output for 4 Objectives and 11 steps per objective Random Weights With random distributed importance factors or weights the Pareto set is sampled randomly if it is convex and the algorithm is not trapped by local minima Random distributed weights are generated Specify number of solutions if random weights are selected WinOpt HDR User Manual 63 Optimization Parameters This dialog is used for the settings of the deterministic algorithm that uses dose variance based objectives WinOpt HDR User Manual 64 Analysis of solutions and decision making tools 1 er AD eee bed ig fe Pemporan Pareto Set D NTissuex 2 gt 1 185 O 8e849 10 593 0 296 Bi objective Temporary Window for the display of the so called bi loss map Displays the distribution of the population of the evolutionary multiobjective optimization algorithm as it evol
95. y implants which include in addition to the surrounding normal tissue maximal three critical structures Version 2 0 is not full optimized A speed up of optimization and dose calculation by a factor of 2 3 is possible The computation time increases linear with the number of active source dwell positions New dose calculation methods could decrease the computation time by an additional factor of 2 5 WinOpt HDR requires that in the registry a directory is specified such as in this example HKEY_LOCAL_MACHINE SOFTWARE Winopt HDR Directory d winopt This directory is expected to exist and that WinOpt HDR is able to create files in this directory The user should modify the path which is this example is set arbitrary to d winopt WinOpt HDR User Manual Definition of Terms OAR CT DVH HDR LDR PTV PC U S VOI Organ At Risk Computed Tomography Dose Volume Histogram High Dose Rate Low Dose Rate Planning Target Volume Personal Computer Ultra Sound Volume Of Interest WinOpt HDR User Manual 8 Introduction In brachytherapy small miniaturized radioactive sources are positioned within the tumor so that an effective dose delivery results while saving healthy tissues This can be done by two principal methods e The sources can be positioned directly into the tumor tissue and will stay there until the end of their life continuously decreasing activity of the sources according to their decay scheme and decay half time T
96. zation Optimization at 10 22 01 11 45 48 Deterministic Optimization Deterministic Multi Objective Weighted Sum Uniform distributed importance factors Sets per objective 11 Number of Objectives 3 Optimization Objectives 1 PTV Surface Variance 2 PTV Volume Variance 3 Volume Variance D gt Dcrit of urethra Using WinOpt HDR Very Fast Deterministic Optimization Algorithm Include OAR urethra with 500 Sampling Points and Dcrit Dref 1 250000 Implant Geometry Reconstruction File Contour File PTV Volume cm 3 52 7220 Surface cm 2 70 2718 Surface Dose Points 208 Dose Points 500 Body Volume cm 3 1213 8017 Surface cm 2 638 8287 Dose Points 500 Rectum Volume cm 3 73 8731 WinOpt HDR User Manual Surface cm 2 105 6765 Dose Points 500 urethra Volume cm 3 2 3627 Surface cm 2 13 1406 Dose Points 500 Surrounding Tissue Dose Points 1000 Number of Catheters 16 Number of Dwells 270 Number of Active Dwells 94 Number of Surface Points 208 Sampling Points for CS or PTV Volumetry 50000 Sampling Points for Body Volumetry 100000 Sampling points 0 950000 mm from Catheter center ignored Here log of another optimization run WinOpt HDR v 2 00 New Optimization Optimization at 10 22 01 11 50 37 Deterministic Optimization Deterministic Multi Objective Weighted Sum Uniform distributed importance factors Sets per objective 11 Number of Obje

PDF file

Contents

Download Pdf Manuals

Related Search

Related Contents