The DSCoMAN User manual is available here.
Contents
1. Baseline Time 25 602 LastTime 58 735 Slice Locations 1 00 Time Points 40 0000 Echo Time msec 32 00 Whole Brain Threshold Noise SD s i 0 00 delta R2 curve to use Raw delta R2 e Parameters for Last baseline Time Computation Last Baseline Time Computation Method Method 3 Use threshold percent of the area under the delta R2 curve v Threshold number of noise standard deviations 13 0 Threshold ofthe area under the delta R2 curve 15 0 Show following Images Whole Brain Mask P Delta R2 l Corrected Delta R2 rCBY V rMTT V rCBF V TTP OK Cancel Figure 3 DSComan 1 0 Raw Perfusion Control Panel 4 1 2 1 Map Selection In Figure 3 you can use the checkboxes to choose which images maps should be generated This section will describe the images that can be generated by the Raw Perfusion plugin This will also help in understanding some of the numeric parameters in the control panel in Figure 3 Whole Brain Mask This is the first step before starting any actual analysis We compute a whole brain mask for each slice in the dataset so that only locations with an average baseline signal intensity more than x times the standard deviation of background noise are included Here x specified by the user using the input parameter Whole Brain Threshold Noise SD s Background noise standard deviation is calculated using the user defined background ROI Further analysis is done only on voxels that are present in the2. Francisco Calif Berkeley Society of Magnetic Resonance 1994 279 dblab duhs duke edu 13 13 Version 1 0 12 18 2006
3. the area under the delta R2 curve of the total area Threshold number of noise standard deviation This parameter specifies the number of noise standard deviations to be used for computation of the last baseline time using Method 2 described above Say the value of this parameter is x Then the last baseline time for a signal intensity curve is the time at which the signal intensity goes x SD s below the average baseline signal intensity Threshold of the area under the delta R2 curve This parameter specifies a threshold to use for the area under the delta R2 curve for the computation of the last baseline time using Method 3 described above 4 1 2 3 Limitations The limitations associated with mean transit time and cerebral blood flow measurements generated with the Raw Perfusion plugin should be carefully considered This plugin uses first moment of the uncorrected delta R2 vs time curve to calculate mean transit time This approximation as has been pointed out in several publications provides at best a rough estimate of relative mean transit time Since the mean transit times generated by this technique are used to compute cerebral blood flow similar precautions apply to the blood flow maps The cerebral blood volumes calculated should also be considered relative not absolute measurements Improved quantitation of mean transit time and flow could be obtained by fitting the underlying curves to a gamma variate equations and or de
4. user specified parameter and is explained in section 4 1 1 2 4 1 2 2 Input parameters This section describes the numeric parameters in the input dialog show in Figure 3 If values are available Raw Perfusion software reads default values of these parameters from the image header Exclude first n timepoints In the DSC perfusion image the first few time points may have very high peaks These points can be excluded from the analysis by specifying the number of initial time points to exclude Last Baseline Time This is the last time in seconds before the gadolinium contrast appears in the venous region of interest See section 4 1 2 for help in calculating the last baseline time This value is used to decide the number of time points to use to calculate the average baseline signal intensity Last Time This is the last time in seconds to be included in the analysis Slice Locations The total number of slice locations in the input image Time Points The total number of time points in the input image Echo Time msec The echo time in msec of the dynamic susceptibility contrast enhanced MR sequence Whole Brain Threshold Noise SD s This is used to define the whole brain mask If you specify a value x for this parameter then the whole brain mask consists of voxels with average baseline signal intensity exceeding x SD s of the background noise delta R2 curve to use This is a popup menu to choose the delta R2 curv
5. whole brain mask Delta R2 This represents the relaxivity time curve a parameter related to the concentration of gadolinium in the voxel It is computed as 1 TE In S t So where TE is the echo time S t is the dynamic signal intensity and So is the average baseline signal intensity Corrected Delta R2 This is the delta R2 curve corrected for contrast agent leakage according to the equation A11 in the Appendix of 1 This computation involves fitting a curve to the data to determine dblab duhs duke edu 9 13 Version 1 0 12 18 2006 the parameters K1 and K2 as described in 1 rCBV This is the relative Cerebral Volume map It is computed from the trapezoidal integration of the deltaR2 map from the first time after the last baseline time to the last time specified by the user That is this integration is done over the part of the deltaR2 curve after the arrival of contrast The deltaR2 map used depends on the choice made in the input parameter delta R2 curve to use rMTT This is the relative Mean Transit Time map The relative Mean Transit Time is defined in equation A1 in the Appendix rCBF This is the relative Cerebral Blood Flow map The Cerebral Blood Flow is defined as the ratio of the relative Cerebral Blood Volume to the relative Mean Transit Time TTP This is the Time to Peak map The Time to Peak TTP is calculated as the time taken for the delta R2 curve to peak from the Last Baseline Time The Last Baseline Time is a
6. DSCoMAN A User s guide to DSCoMAN DSCoMAN version 1 0 1 Introduction Although imaging of perfusion using T or T weighted MR images called dynamic susceptibility MR imaging has been performed for many years the methods used to analyze these images are as a rule far from widely accepted The use of this type of imaging while promising is arguably far from validated for any particular clinical application One barrier to the validation of these techniques is that there is a lack of widely available reference methods to analyze these images In the absence of comparable acquisition and analysis methods it is difficult to directly compare quantitative results obtained in one laboratory with that of another The function of a reference technique is not necessarily to provide the ideal analysis scheme but to provide a reasonable method that can be easily applied for the purposes of comparison As a first step towards providing a reference technique particularly for the application of dynamic susceptibility MR imaging to brain tumors we have developed the DSCoMAN software package DSCoMAN stands for Dynamic Susceptibility Contrast MR ANalysis This software implements a method of dynamic susceptibility contrast DSC MR analysis that takes into account that the blood brain barrier is disrupted in patients with high grade gliomas When there is a significant blood brain barrier breakdown then the relative cerebral blood volume rCBV
7. ation of gadolinium in the voxel It is computed as 1 TE In S t So where TE is the echo time S t is the dblab duhs duke edu 5 13 Version 1 0 12 18 2006 dynamic signal intensity and Sois the average baseline signal intensity K1 This is the map of the parameter K1 as defined in equation 1 in 1 K2 This is the map of the parameter K2 as defined in equation 1 in 1 The K2 term reflects the effects of contrast agent leakage Rsquare This is the map of R for the linear fit of the model in equation 1 in 1 rCBV Uncorrected This is the relative Cerebral Blood volume map that does not take into account contrast agent leakage rCBV Corrected This is the relative Cerebral Blood volume map corrected for contrast agent leakage It is defined in equation 2 in 1 4 1 1 2 Input parameters This section describes the numeric parameters in the input dialog show in Figure 2 If values are available Boxerman Weisskoff software reads default values of these parameters from the image header Exclude first n timepoints In the DSC perfusion image the first few time points may have very high peaks These points can be excluded from the analysis by specifying the number of initial time points to exclude Last Baseline Time This is the last time in seconds before the gadolinium contrast appears in the venous region of interest See section 4 1 2 for help in calculating the last baseline time This value is used to decide the number of
8. convolution of the tissues curves using an arterial input function but these techniques are not offered in this version of the DSCoMAN software 4 1 3ROI TimeCourse When you click on the ROI TimeCourse button in Figure 1 the ROI TimeCourse plugin is invoked This plugin can be used to see the variation in signal intensity over time The ROI TimeCourse plugin is part of the TOPPCAT software package The download instructions are available at http dblab duhs duke edu modules dblabs_topcat index php id 1 For a region of interest on an image ROI Time Course creates a plot window as seen in Figure 4 showing the variation in the mean signal intensity within the location specified by the region of interest dblab duhs duke edu 11 13 Version 1 0 12 18 2006 over all time points This is very useful for calculating the proper last baseline time point before the arrival of the contrast bolus In the plot the x axis is designated in seconds and the y axis is signal intensity Note placing a cursor over the plot window will help indicate the time in seconds corresponding to the location of the cursor Selecting the List button generates a list the coordinates of the points on the curve lox 670x250 pixels 8 bit 139K n c a E 5 Y Time secs List save Copy Fig 4 Sample output from the ROI time course plugin 4 1 4 HyperVolume When you click on the HyperVolume button in Figure 1 the HyperVolume plu
9. e that should be used for the TTP rMTT and rCBF rCBF and rCBV calculations You can choose the raw delta R2 curve or the corrected delta R2 curve from a Boxerman Weisskoff analysis dblab duhs duke edu 10 13 Version 1 0 12 18 2006 Last Baseline Time Computation Method This is a pop up menu to choose the method to use for computing the last baseline time The possibilities are 1 Method 1 Use user specified Last Baseline Time This method uses the Last Baseline Time specified by the user in Figure 3 It uses the same value of the last baseline time for each pixel 2 Method 2 Use threshold SD below baseline intensity This method uses a threshold SD to compute the last baseline time individually for each pixel In this case the last baseline time is identified when the signal intensity goes below the specified SD s specified by input parameter Threshold number of noise standard deviations of the baseline signal intensity Note that this computation is done on a pixel to pixel basis 3 Method 3 Use threshold percent of the area under the delta R2 curve This method uses the area under the delta R2 curve to compute the last baseline time for each pixel For a given pixel it computes the area under the delta R2 curve at each time point It identifies the last baseline time as the point at which the area under the curve at a particular time exceeds a certain user specified threshold specified by input parameter Threshold of
10. ed ImageTimeTable txt containing the time for each time point However if this information is not available in the image header then the user must supply the ImageTimeTable txt file containing the time for each time point one time per line 3 When you click on the Raw Perfusion menu in Figure 1 the Raw Perfusion control panel shown in Figure 3 should appear This control panel allows the user to provide various input parameters It also allows the user to select the maps that should be displayed Section 4 1 2 1 describes the different images that can be generated using the Raw Perfusion plugin Section 4 1 2 2 describes the numeric input parameters in Figure 3 dblab duhs duke edu 7 13 Version 1 0 12 18 2006 DSCOMAN ersion 1 0 Exclude first n timepoints Last Baseline Time 25 603 LastTime 58 735 Slice Locations 16 00 Time Points 40 0000 Echo Time msec 32 00 Whole Brain Threshold Noise SD s 10 00 delta R2 curve to use Raw delta R2 Parameters for Last baseline Time Computation Last Baseline Time Computation Method Method 3 Use threshold percent ofthe area under the delta R2 curve 7 Threshold number of noise standard deviations 3 0 Threshold ofthe area under the delta R2 curve 5 0 Show following Images Whole Brain Mask J Delta R2 l Corrected Delta R2 M rMTT V rCBF MV TTP dblab duhs duke edu 8 13 Version 1 0 12 18 2006 DSCoMAN 1 0 Raw Perfusion x Exclude first n timepoints Last
11. f you already have TOPPCAT in your ImageJ plugins folder then you do not need to put these class files into your DSCoMAN folder ImageJ will give a Duplicate Command warning if you have multiple classes with the same name in your plugins folder This version of DSCoMAN depends on other utility classes that can be obtained by clicking the link for Download the Utilities on the DSCoMAN webpage Unzip the file called DBLAB _1 0 zip and save it into the main ImageJ plugins folder i e ImageJ_ Home plugins where Image Home is the home directory of your ImageJ installation This will save the DBLAB Jar into the main plugins folder Note If you had a previous version of DBLAB jar in your Image plugins folder overwrite it with the latest DBLAB jar available on the DSCoMAN webpage Now restart ImageJ You should see three different items under DSCoMAN in the plugins menu Boxerman Weisskoff DSCoMAN GUI and Raw Perfusion The DSCoOMAN GUI is a user interface for calling other plugins The Boxerman Weisskoff plugin generates the corrected rCBV map using the methods described in 1 The Raw Perfusion plugin generates maps of parameters such as Time To Peak TTP relative Mean Transit Time rMTT and relative Cerebral Blood Flow rCBF DSCoMAN Operation DSCoMAN can be used either in manual mode or in macro mode To get familiar with the DSCoMAN software and to perform small scale analyses we suggest using manual mode Macro mode is usefu
12. gin is invoked This plu gin can be used to display a four dimensional xyzt dynamic stack with two slider bars for easy manip ulation in both the time and z dimensions The HyperVolume plugin is part of the TOPPCAT software package The download instructions are available at http dblab duhs duke edu modules dblabs_topcat index php id 1 HyperVolume makes use of the most excellent HyperVolume Browser plugin from Patrick Pirrotte and Jerome Mutterer rsb info nih gov 1j plugins hypervolume browser html 4 2 Macro mode The DSCoMAN software can be operated using ImageJ s macro language This is very useful when you need to perform a number of repetitive analyses To experiment with macro language use the ImageJ command Plugins Macros Record and or refer to the ImageJ website One disadvantage to using the macro mode is that you must explicitly supply all parameters within the macro text Specifically the parameters supplied in the macro text override default parameters It is possible however to modify the macro text to supply these parameters in an automated way by using our Query Dicom Header plugin rsb info nih gov 1j plugins query header html within the macro to make the macro text file aware of the Dicom parameters used in the stacks dblab duhs duke edu 12 13 Version 1 0 12 18 2006 Here is a sample ImageJ macro code to run the Boxerman Weisskoff plugin It provides the required input parameters and specifies that the follow
13. ime point in order to perform the analysis If this information is available in the image header then on clicking on Boxerman Weisskoff in the menu in Figure 1 it will automatically read the image times and display a file called ImageTimeTable txt containing the time for each time point However if this information is not available in the image header then the user must supply the ImageTimeTable txt file containing the time for each time point one time per line When you click on the Boxerman Weisskoff menu in Figure 1 the Boxerman Weisskoff control panel shown in Figure 2 should appear This control panel allows the user to provide various input parameters It also allows the user to select the maps that should be displayed Section 4 1 1 1 describes the different images that can be generated using Boxerman Weisskoff Section 4 1 1 2 describes the numeric input parameters in Figure 2 Version 1 0 12 18 2006 DSCoMAN 1 0 Boxerman Weisskoff Exclude first n timepoints Last Baseline Time Last Time Slice Locations Time Points Echo Timet msec Whole Brain Threshold Noise SD s Non Enhancing Pixels Threshold SDs above below baseline or sar Mo Tri a m a Non Enhancing Pixels detected on final n timepoints Show following Images F Whole Brain Mask Delta R2 l Non enhancing Pixel Mask T K1 V K2 l Rsquare V rCBY Uncorrected V rCBY Corrected OK Cancel Figure 2 DSCoMAN 1 0 Boxerman Weisskoff Control Pane
14. ing maps be generated K1 K2 Rsquare rCcBV UnCorrected and rCBV Corrected This is specified by the boolean parameters k1 k2 rsquare rcbv_uncorrected and rcbv_corrected in the input parameter list run Boxerman Weisskoff exclude_first_n_timepoints 1 last_baseline_time 25 603 last_time 58 735 slice_locations 16 time_points 40 0000 echo_time msec 32 whole _ brain threshold noise_sd s 10 non enhancing pixels threshold sds _above below_baseline 1 non enhancing pixels detected_on_final_n_timepoints 10 k1 k2 rsquare rcbv_uncorrected rcbv_corrected Appendix The rMTT for relative mean transit time can be estimated as the first moment of the delta R2 curve joa RD Al rMIT A R2 t 0 This integration is done from the first time after the last baseline time to the last time specified by the user That is this integration is done over the part of the curve after the arrival of contrast The initial baseline portion of the curve is ignored References 1 Boxerman JL Schmainda KM Weisskoff RM Relative Cerebral Blood Volume Maps Corrected for Contrast Agent Extravasation Significantly Correlate with Glioma Tumor Grade Whereas Uncorrected Maps do not AJNR Am J Neuroradiol 2006 27 859 67 2 Weisskoff RM Boxerman JL Sorensen AG et al Simultaneous blood volume and permeability mapping using a single Gd based contrast injection Proceedings of the Society of Magnetic Resonance Second Annual Meeting 1994 Aug 6 12 San
15. l 4 1 1 1 Map selection In Figure 2 you can use the checkboxes to choose which images maps should be generated This section will describe the images that can be generated by the Boxerman Weisskoff plugin This will also help in understanding some of the numeric parameters in the control panel in Figure 2 Whole Brain Mask This is the first step before starting any actual analysis We compute a whole brain mask for each slice in the dataset so that only locations with an average baseline signal intensity more than x times the standard deviation of background noise are included Here x specified by the user using the input parameter Whole Brain Threshold Noise SD s Background noise standard deviation is calculated using the user defined background ROI Further analysis is done only on voxels that are present in the whole brain mask Non enhancing Pixel Mask This mask is computed for each slice in the dataset and consists of those voxels that did not demonstrate signal intensity enhancement averaged over the final x time points greater than y standard deviation above that pixel s average baseline intensity Here x and y are user specified x specified by the input parameter Non Enhancing Pixels detected on final n timepoints and y is specified using the input parameter Non Enhancing Pixels Threshold SDs above below_baseline Delta R2 This represents the relaxivity time curve a parameter related to the concentr
16. l when you need to perform a number of repetitive analyses 4 1 Manual mode The easiest way to use manual mode is to click on the ImageJ plugins menu and then click on the DSCoMAN GUI plugin This will show an interface as seen in Figure 1 dblab duhs duke edu 3 13 Version 1 0 12 18 2006 DSCoMAN 1 0 Raw Perfusion ROI TimeCourse Hypervolume yzi Close All Windows Figure 1 DSCoMAN 1 0 GUI Using ImageJ open the dynamic susceptibility contrast MR images that you want to analyze as a stack of images You can then use the menu items on the DSCoMAN GUI shown in Figure 1 to perform different functions as described in this section 4 1 1 Boxerman Weisskoff The Boxerman Weisskoff plugin generates the corrected rCBV map using the methods described in 1 To perform Boxerman Weisskoff analysis on a set of dynamic susceptibility contrast MR images l 2 4 Click on the OK button to activate the plugin dblab duhs duke edu 4 13 Open the dynamic susceptibility contrast MR images as a stack in ImageJ for example use File Import Image Sequence Place a region of interest in the background of the image and then click on Boxerman Weisskoff in the menu in Figure 1 The background ROI is required for computing the whole brain mask Boxerman Weisskoff analysis is only performed on whole brain voxels defined in Note that the Boxerman Weisskoff plugin requires the information about the image times at each t
17. obtained by trapezoidal integration of concentration time curves underestimates true cerebral blood volume due to contrast agent leakage 1 The DSCoMAN software implements the method described by Boxerman et al 1 The cerebral blood volume maps corrected using this method were recently shown to more accurately correlate with tumor grade in patients with glioma than uncorrected methods The DSCoMAN software generates the corrected relative Cerebral Blood Volume maps from DSC perfusion images It can also generate maps of K2 a measurement of contrast agent leakage In addition DSCoMAN can be used to create Time to Peak TTP maps as well as estimates of relative Mean Transit Time rMTT and the relative Cerebral Blood Flow rCBF 2 What is needed to run DSCoMAN dblab duhs duke edu 1 13 Version 1 0 12 18 2006 2 1 Hardware requirement DSCoMAN is designed to run on any hardware running an operating system that supports ImageJ a public domain open source Java based image processing program authored and maintained by Wayne Rasband at the National Institute of Mental Health The operating systems supported by ImageJ include Mac Windows and Linux platforms 2 2 Software requirement DSCoMAN has been implemented as plugins to ImageJ ImageJ rsb info nih gov ij must be installed before the DSCoMAN plugins We recommend installing the version 1 36 or a later version of ImageJ bundled with Java 1 5 This can be obtained from http rsb info nih go
18. tallation procedure 3 1 Install ImageJ If you have not already installed ImageJ download and install this free program by following the instructions at rsb info nih gov 1j download html We recommend installing the version 1 36 or a later version of ImageJ bundled with Java 1 5 The DCSoMAN software requires Java 1 5 3 2 Adjust Memory In order to handle large sets of images maximize the RAM available to ImageJ Use the Edit Options Memory command to make more than the default 128MB available to ImageJ It often helps to increase the memory allocation option to approximately 2 3 of available RAM e g mx170m on a 256MB machine Larger increases may lead to memory thrashing Another way to make more memory available to ImageJ is by running from the command line and using the Xmx option Note that dblab duhs duke edu 2 13 Version 1 0 12 18 2006 ImageJ is limited to 64MB when you run it by double clicking on ij jar 3 3 Install the DSCoMAN plugins Follow these steps 4 Make a directory called DSCoMAN in your ImageJ plugins folder Unzip the DSCoMAN software and save it in the DSCoMAN directory in your plugins folder The zip file will contain a jar file called DSCoMAN_1 0 jar containing the DSCoMAN classes The zip file will also contain class files for the ROITimeCourse and the HyperVolume plugins used by the DSCoMAN software Note The ROITimeCourse and the HyperVolume class files are also a part of the TOPPCAT package I
19. time points to use to calculate the average baseline signal intensity The Time to Peak TTP is calculated as the time taken for the curve to peak from the last baseline time Last Time This is the last time in seconds to be included in the analysis Slice Locations The total number of slice locations in the input image Time Points The total number of time points in the input image Echo Time msec The echo time in msec of the dynamic susceptibility contrast enhanced MR sequence Whole Brain Threshold Noise SD s This is used to define the whole brain mask If you specify a value x for this parameter then the whole brain mask consists of voxels with average baseline signal intensity exceeding x SD s of the background noise Non Enhancing Pixels Threshold SDs above below_baseline This is used to define the Non enhancing pixel mask If you specify a value x for this parameter then the Non enhancing pixel mask consists of all the pixels that did not demonstrate signal intensity enhancement greater than x SD s above that pixel s average baseline signal intensity dblab duhs duke edu 6 13 Version 1 0 12 18 2006 Non Enhancing Pixels detected on final n timepoints This parameter specifies the number of tail time points to use while computing the signal intensity enhancement for generation of the Non enhancing pixel mask For example if the parameter for Non Enhancing Pixels detected on final n timepoin
20. ts has value x and the parameter for Non Enhancing Pixels Threshold SDs above below_baseline has value y then the Non enhancing pixel mask consists of all the pixels that did not demonstrate signal intensity enhancement signal intensity averaged over the final x time points greater than y SD s above that pixel s average baseline signal intensity 4 1 2 Raw Perfusion When you click on the Raw Perfusion button in Figure 1 the Raw Perfusion plugin is invoked This plugin can be used for generating maps of parameters such as Time To Peak TTP relative Mean Transit Time rMTT and relative Cerebral Blood Flow rCBF To perform the Raw Perfusion analysis on dynamic susceptibility contrast MR images 1 Open the dynamic susceptibility contrast MR images as a stack in ImageJ for example use File Import Image Sequence 2 Place a region of interest in the background of the image and then click on Raw Perfusion in the menu in Figure 1 The background region of interest is required for computing the whole brain mask The Raw Perfusion plugin does analysis on the whole brain voxels defined in this mask Note that the Raw Perfusion plugin requires the information about the image times at each time point in order to perform the analysis If this information is available in the image header then on clicking on Raw Perfusion in the menu in Figure 1 it will automatically read the image times and display a file call
21. v 1j download html 2 3 MR Imaging data DSCoMAN software uses images acquired from dynamic susceptibility MR imaging as input Most frequently we expect that these images will be available as DICOM images although any 8 bit or 16 bit image format that can be opened in ImageJ could be used for example uncompressed tif jpg or Analyze image One advantage of using DICOM images is that DSCoMAN will take default values of parameters such as slice locations echo time etc from the DICOM header This software considers the opened dynamic susceptibility MR imaging set to be a 4D stack of images in xy zt order composed of images repeated at the same number of image locations n for a certain number of time points n In fact the stack of images in the image set should be composed of image width image height ni n pixels It is important that some of the early time points be obtained before the contrast agent is administered to establish the baseline signal intensity The data obtained earliest should begin the stack with subsets proceeding in order of increasing time so that the data obtained last ends the stack HINT If your stack is in xy tz order Michael Abramoff has provided the handy Hypervolume Shuffler rsb info nih gov 1j plugins hypervolume shuffler html that can place these into xy zt order For the purposes of making stacks it certainly helps to have all of the relevant images in a single directory 3 Ins
Download Pdf Manuals
Related Search