The SIMBA Image Management and Analysis System - VIA
Contents
1. 36335 55 0 6797 x 0 6797 x 3 75 n5 Apr 25 2000 42 94 34 24 2101 1 93 0 6816 x 0 6816 x 2 50 Growth Information Date T1 Date T2 A change PT e n1 and n3 are different n1 n3 Jan 12000 Feb 15 2000 45 48 30 47 27 35 6 resolutions e ni and n5 are different ni n5 Jan 12000 Apr 25 2000 115 70 10 66 02 27 0 resolutions e n3 and n5 are different n3 n5 Feb 15 2000 Apr 25 2000 70 42 17 88 61 20 9 resolutions Figure 10 Growth analysis of three scans of the nodule shown in Figure 8 20 Growth Analysis Mozilla F File Edit View History Bookmarks Tools Help Axial View Jan 1 2008 Feb 15 2686 Apr 25 2000 Coronal View Jan 1 2686 2688 Apr 25 2686 Figure 11 Axial and Coronal visualizations of the progression of the nodule shown in Figure 6 based on the manually markings of the radiologist 4 Computer Assisted Marking In this experimental method the marker and the computer work together At certain times the computer may suggestions for the location of a boundary segment and the marker may accept reject or modify these suggestions The main benefits of this approach are to improve the performance of the reader by creating boundaries that are more repeatable and require less effort than manual marking Furthermore the computer assisted method can be guided by the marker in complex situations that cannot be correctly resolved by the fully automated marking methods a Perpendicular2. Measure b Measuring Grid Figure 12 Screen snapshots of the standard measuring tools 21 5 Standard measuring tools n addition to the advanced measuring methods the image viewer provides the standard electronic calipers and measuring grids that are typical of conventional medical image display devices see Figure 12 a and b for examples Measurements made by these conventional techniques may be compared with the more detailed area and volumetric measures 6 Additional CT lung analysis tools The system is designed to accept any number of additional computer aided diagnostic tools as they become available As we proceed we envision the development of an entire suite of new tools for tumor response assessment Currently two image analysis tools in addition to growth assessment have been installed automated nodule detection and emphysema quantification The automated pulmonary nodule detection system identifies candidate locations for nodules in a whole lung CT scan and the user reviews these locations selects a subset of these candidates for growth analysis For emphysema detection a standard image density algorithm is used to estimate the degree of emphysema throughout the lung region Quantitative results including a measure of lung volume and emphysema index and a coronal image representation of the distribution of the emphysema are provided by the web system A snapshot from the nodule detection tool is shown in Figure 13 and a snapsh
3. back to the site The center has no access to patient identification information and the site is entirely responsible for the care of its subjects This center monitoring provides for exceptional QA for all the different studies and for pooling of the resulting study data The intelligent data forms are highly optimized for lung cancer screening Based on over 10 years of ELCAP experience they have received numerous revisions Furthermore they involve a very high degree of self checking before submission is permitted Therefore incomplete data cannot be entered into the system A pending form facility permits the temporary storage of incomplete forms prior to submission to the database Data forms have been developed for all aspects of study operation including intake background CT image reading X ray reading PET reading biopsy cytology report histology report pre intervention intervention and follow up Some of the key features of this system are as follows 1 Built in data entry checking data fields are checked for completeness valid values and logical consistency before they are submitted If problems are detected then they must be corrected before the forms will be accepted by the system This mechanism provides timely feedback to the user and prevents incomplete or inconsistent data from being entered into the system 2 A pending form facility is available for data forms that for which all needed information is mot immediately avai
4. described in detail in Appendix A Images are managed in collections which are called cases The meaning of a case is project dependent in the context of a clinical study a case is typically a subject From the home page one can select a specific case of a subset of cases Example of a subset of cases is shown in Figure 1 pu File Edit View History Bookmarks Tools Help PR Case Review List Return Home Hep lcasew Date j item amp ection ssooot lotisanr2000 Analysis with and without spiculation mage List ssoo03 lotwani2000 mass abuts mediastinum Region Analysis Image List ssooos lotwsanr2000 mulitiobulated mass image List ssoot0 loti4anr2000 irregular shape Image List ssoo11 lotisanr2000 Bxo0 9 Resion ana mage List lssoot2 01 Jan 2000 large complicated mass image List Figure 1 example case list From the case list the options are in this case to go directly to the data analysis for a selected case or to review a single case from the list An example of selecting a case from the list or equivalently specifying a single case in the home page is given in Figure 2 Sa Ean i Ts O File Edit View History Bookmarks Tools Help PR Case Review for 550023 All Case Review hae Case ID Date Item Action ssoo23 lo1Wan 2000 plaque like pleural mass Image Annotations Code r Time Senes Acquisition Identifier x Action bBX0037 133211 00 02
5. make modifications to ensure that the judgments for ambiguous boundary regions may be treated consistently When multiple ROIs are available a growth analysis similar to that for the fully automated method may be performed An Example of the area method is illustrated in Figure 4 Once the user has marked the areas of interest in the image they are shown side by side in the web viewer as shown in Figure 5 This permits the user to make adjustments to ensure that exactly the same region is being selected in each image E 7 ont x Growth Analysis Mozilla Firefox N C File Edit View History Bookmarks Tools Help a Detailed Nodule View Showcurrentgroups 3D auto 3D manual Return Home her g Group 2 2D manual analysis Nodule Information Size mm Size mm Area Volume A i a aa from X Y Extent from Area mm mm a ig n3 Jui 29 1999 11 18 10 29 83 20 570 86 0 1875 x 0 1875 x 1 25 Edit n5 Aug 30 1999 12 49 11 53 104 39 802 38 0 1875 x 0 1875 x 1 00 Edit Growth Information Nodules Date T1 Date T2 interval Notes Days aan Change Jul 29 1999 Aug 30 1999 40 56 65 09 Image Region of Interest Figure 4 Growth analysis web snapshot showing the area growth analysis for two time separated scans Note the scans are taken at different resolutions one prone and one supine and only the central most slice is considered 17 Growth Analysis Mo
6. method is similar to that of the manual area method with the difference that the boundary on each image that the lesion is visible is marked and a volumetric rather than an area based growth estimate is performed 4 Standard calipers and measuring grids In addition to the advanced measuring methods the image viewer provides the standard electronic calipers and measuring grids that are typical of conventional medical image display devices both RECIST uni dimensional and WHO bi dimensional conventions are supported Measurements made by these conventional techniques may be compared with the more detailed area and volumetric measures 5 Additional tools for CT lung analysis The system is designed to accept any number of additional computer aided diagnostic tools as they become available Currently two such tools have been installed automated nodule detection and emphysema detection 6 Experimental web based infrastructure for image analysis A web based infrastructure is available for the support of new image analysis strategies This supports a number of experimental image based research projects through a simply customized web interface For example customized experimental interfaces have been developed for dynamic cardiac MRI image analysis 3 SIMBA for computer vision research 3 1 Home Page and Case management General research projects employ a generic home page that is configurable for different project requirements This home page is
7. of different locations The primary capability of SIMBA is a documented image archive Researchers may upload images acquired during experiments with full documentation and then retrieve them from the system as needed The system provides a secure repository for all the experimental data including images The second goal of the system is to provide an interactive image review system though the web In this way users can review the image data and associated documentation and make image measurements in the Same web based system All records for a study may be maintained in one secure location All images including hand written notes can be managed by the system The third main goal of the system design was to create an interface between the experimenters and research activities in computer algorithms for image analysis As an example of this consider the growth measurement of pulmonary nodules The algorithm research group is developing and refining automated methods for measuring the size change in lesions over time and hence determining the clinically important measurement of growth rate The system facilitates the identification of lesions by clinical users and their analysis by the computer algorithms Two major benefits of this scheme are first that clinical users get instant access to algorithms updates and second clinical users and algorithm developers can easily interact on a case by case basis as needed 2008 A P Reeves 1 A second ex
8. 3 27 0 1875 x 0 1875 x 1 00 Growth Information Date T1 Date T2 interval siratni Notes Days chane ow Gl nan Jun1320000 13 2000 Ju 192000 19 2000 l 20 84 131 78 ma 3 n10 Jun 13 2000 Sep 26 2000 105 79 44 124 47 3 8 2 nr n10 sul 19 2000 Sep 26 2000 69 43 50 120 96 18 8 Solid Component Figure 3 Growth analysis for the three volume measurements shown in Figure 2 The first table shows the volumes from the three image sets and the second table shows the growth rates computed for all three image pairs 16 2 Manual two dimensional growth analysis In many cases of pulmonary nodules the totally automated methods may produce the most accurate results However for other lesions with less well defined margins and for cases where the scan parameters between studies are significantly different automated methods cannot currently be used For these situations interaction between the radiologist and the system is necessary The radiologist indicates where they judge the location of the image region of interest ROI typically this is a lesion but may be any anatomical feature and the system then performs an analysis on these measurements For the 2D area growth analysis method the radiologist selects the best central image of a lesion and carefully marks the boundary of the ROI When multiple studies have been obtained and ROls have been created the user can compare them side by side and
9. AP projects using a central server model so support lung cancer clinical studies at over 30 institutions worldwide including China Japan and several locations in Europe This has involved over 30 000 subjects 50 000 lung CT scans all major brands of CT scanners are represented as are many different PACS systems In addition to the ELCAP studies SIMBAhas been used to support several other studies in Europe in this case the system was installed on server computers in Europe and is managed by local personnel The system has a high degree of security double password entry is required all communications are encrypted Web based interactions use the secure sockets layer SSL protocol and background computer to computer communications use the secure shell SSH protocol SIMBA for lung cancer studies comprises of three tightly integrated components a clinical trials data management system a comprehensive image archive and a set of on line image analysis tools Each of these will be considered in turn 2 1 Clinical Trials Management The clinical trial management component has a conventional data form basis with the exception that the database is web accessible Access to the system is supported by a two level center site model A single center monitors the activities at a number of sites Each site is an autonomous entity with its only study however data collected by the site is monitored by the experienced center personnel which report any findings
10. The SIMBA Image Management and Analysis System Overview Release 1 November 2008 Anthony P Reeves PhD Abstract The SIMBA Image Management and Analysis System provides an interactive web based facility to meet the needs of research projects that involve image data The main needs for such projects includes image transfer and archiving image documentation case based image management image analysis data from management and project outcome analysis This system has been used to support a diverse set of research projects involving image data SIMBA has two major configurations a a general purpose image analysis system to support image based research projects and b from a collaboration with the Early Lung Cancer Action Program ELCAP of Weill Cornell Medical College a system for managing multi center clinical studies 1 Introduction Research involving image data has specific requirements with respect to the management of large image data files and the analysis of the image content While currently much image analysis is performed by human observation computer algorithms are being increasingly used to facilitate more objective and quantitative analysis methods The System for Image Management extended By Analysis SIMBA has been designed to facilitate image based research across a broad range of applications in addition it is built upon a web based framework to support research groups that may be distributed physically across a number
11. This method is also used in the CRPF funded image database that documents the growth of large pulmonary lesions The growth analysis presentation by both the automated and manual volumetric methods may be reviewed by going to this public database http www via cornell edu crpf html An example of the web snapshot showing a manually marked boundary for all images of a large nodule from the CRPF database is shown in Figure 8 In Figure 9 visualizations of the manual marked nodule region are shown and the corresponding growth analysis is shown in Figure 10 Fle Edt Yew Go Bookmarks Tools Help s 79 hitps j wonko vis cornel edujeg bin datacjnpreview co index pg fs Nodule Analysis PROOOL 1 Sat Jan 01 10 03 35 2000 Volume 70275 88 mm 3 Approx Size from Volume 1 20 mm Series 02 Acquisition 01 Image 14 at 306 251 Return to Case Rewiew Figure 8 Screen snapshot of the manually marked boundary of a large pulmonary nodule 19 Region Analysis Preview Mozilla F File Edit View History Bookmarks Tools Help Growth Analysis Mozilla Firef File Edit View History Bookmarks Tools Help ESS Growth Analysis for SS0001 Baa Mame Detailed Nodule View Show current groups hee Group 1 3D manual analysis Nodule Information Date Ta Sizo mm Volume Resolution from X Y Extent from Volume mm n1 Jan 1 2000 62 37 51 20 70275 88 0 6250 x 0 6250 x 3 75 n3 Feb 15 2000 50 44 41 09
12. age viewer Information 4 Click on Return Home When done Boundary Marking 1 Select an image to view as outlined above Click on the E putton for instructions on how to do a Boundary Marking marking 2 Click on the Hide button next to the Reference Bnd to remove any reference markings Helo optional 3 Click on Save As button to record your boundary marking Figure 3 Generic Home Page 13 The home page provides access to the system features and also includes yellow help buttons that provide on line access to instructions for using the system Therefore for most projects no additional information is needed beyond access to the home page An exception to this is the clinical studies program in which formal training should be undertaken before using the system The generic home page illustrated in Figure 3 contains the following 1 Project Title The top of the page includes the project title and the page function name for all pages in the system 2 The system message This is the message response from the last action that was performed by the system In Figure 3 Welcome so the response from a correct log on 3 Project Specific Information Projects are customizable if a project contains additional capabilities or requires additional actions this is outlined in this section 4 Main menu buttons This menu contains the main actions that are typically performed including project documentation an result a
13. ample is the tracking over time of vessel diameters in an intravital video image sequence The standard method to accomplish this task is to have a researcher view the image sequence while manipulating electronic calipers to measure the vessel diameter in each image frame A custom designed computer algorithm was developed for this task and integrated into SIMBA Now the user uploads an image sequence marks the vessel diameter in the first image frame and the system then automatically tracks the diameter thought the whole image sequences and provides the user with an analysis of the tracked trajectory The user interacts with the system to review and approve the tracking results and to make any required mid course corrections This example takes advantage of the project customizable capabilities of the system In the following sections a description of the main capabilities of SIMBA are presented the foundation capabilities of the system are outlined the remainder of section 1 An initial application for SIMBA was the management of a multi center clinical study involving medical CT image datasets For this study many additional features such as subject scheduling and study protocol support were added to the system so that all study activities could be managed in a single web based entity This application is described in section 2 Finally the facilities for general image research and customizable projects are outlined in section 3 Appendix A provides in
14. e image analysis function on any image modality All of the image analysis tools we describe are available directly on the web system and do not need to be performed on a separate workstation In addition to the general purpose tools we have also developed some application specific tools for the analysis of chest CT scans These tools include the detection and characterization of pulmonary nodules and the quantification of emphysema The currently available image analysis tools are as follows 1 Automated three dimensional volumetric nodule growth analysis 2 Manual two dimensional lesion growth analysis a Boundary marking b RECIST uni dimensional measure c WHO bi dimensional measure Manual marked volume lesion growth analysis Computer Assisted volume measurement Standard Calipers and measuring grids Whole lung nodule detection and emphysema analysis Se eae 1 Fully automated three dimensional volumetric nodule growth analysis A user requests a nodule analysis by simply clicking with the mouse on the image of a nodule displayed on a web browser with the java viewing tool outlined in the previous section The automated system then determines Preview Mozila metor egion Anal Fi Edn Wew Hiem Bockmaks Tool Help Growth Analysis Retan to Case Review EE 3 Tue Jun 13 10 21 21 2000 Volume 336 19 mm 3 Equiv Diameter 8 63 mm Series 03 ACQuismonc 01 image 13 at 227 243 Figure 1 Screen snapshot Once a nodule has be
15. e presented with the project home page You should never share your personal passkey with others only one session can be open at a time for each personal passkey 12 2 Project Home page Projects may have individual customized home pages however most projects use the generic home page which is shown in Figure 3 or a variation of that home page m Mozilla Firefox File Edit View History Bookmarks Tools Help TR Update review Testbed Home new home testbed Project Specific This is an experiment validation account Information Case Review and Data Entry Image Management System a Main Menu a Locate by View All Cases Case ID Buttons i View Project Reports eee Sea eos Ja Detailed Control User Preferences REONE DO Panel Sepa View all Cases Help for Main Activities Case Selection Region and Change Analysis DICOM Image Download Case management Image Upload Case Management if permitted VI A Vision and Image Ata YSIS GTONP Project Options Management System Updates Password Image Viewing SS Management 1 To select a case either click on View All Cases then click on the Image List button for the desired case or enter the case number in the Case ID field and then click on the associated Go button 2 Click on the View button below the image of interest System Help 3 Click on the L l _ putton in the image viewer for assistance on using the im
16. ed images have associated document pages in which ongoing archival notations may be made Customizable Image Analysis An application programming interface API and related web infrastructure is available that permits the inclusion of project specific image analysis functions without compromising the system security Standard to the API are project documentation image analysis and report generation 6 Archival Messaging There is a built in messaging system that permits users to make annotations and share them with other users Users may also respond to messages as in email All messages and responses are maintained in an archival system as a permanent experiment record 1 3 SIMBA Implementation The SIMBA system is implemented with the apache web server on a Linux computer system SIMBA is distributed as two software components the SIMBA web system which is a set of apache server scripts and support functions and the v4 image processing system that handles all the low level image manipulation operations 2 SIMBA for Lung Cancer Screening Studies SIMBA for the ELCAP was developed to support clinical studies in Lung cancer screening and to provide a data pooling capability to combine the data from different lung cancer studies The novel features of the SIMBA system are that it provides comprehensive support for all aspects of the clinical study and the all functions are accessible via the web The system has been thoroughly tested for the ELC
17. en selected the system presents the nodule ROI left and a montage of images through the nodule right 15 the volume of the nodule In the case of sub solid nodules part solid or nonsolid GGO the system computes a volume for both the solid and nonsolid nodule components Once volumes for a nodule from multiple scans are available a growth analysis can be performed The system reports on growth rates obtained by the size change for all possible pairs of time separated scans of a nodule The automated measuring tool is illustrated in Figures 1 to 3 Figure 1 shows a web display of the nodule once the user has selected it Many other visualizations are available Figure 2 shows a comparative visualization of the same nodule from three time separated scans and Figure 3 shows a snapshot of the growth analysis for this case File Edit View History Bookmarks Tools Help Axial View Jun 13 2008 Jul 19 2666 Sep 26 2608 m T1 T2 T3 Figure 2 Web snapshot of a nodule at three sequential times T1 T2 and T3 axial light shaded visualizations coronal and sagittal views are also available e ileal r Detailed Nodule View show current groups 3D manual He Group 1 3D automated analysis Nodule Information 2 mm 2 mm iE eae ero yeees oe a _ 13 2000 fg 8 3 63 336 19 0 1875 x 0 1875 x 1 00 sul 19 2000 9 19 406 25 0 1675 x 0 1875 x 1 00 nig Sep 26 2000 11 63 10 48 60
18. ettings Jan D1 2000 10008035 insge 1105 D625 x 062i sre Feb 15 2000 05 33 45 image 1500 O 68men 0 6 T l gt Wamms vindia 400 j Fomiary goloct Focus VSW 400 z p i _s0Wwer Focus No Wegeurer Fae __ Sewer Focus No Weseurer os anal g Eqitpannl s Zoomin Zoom out d Ale Grid x a Zohar Curt A No Grid w Pan fange w becin Aod eS Pan Singhs w paua CH Figure 15 Snapshot of web based java viewer showing the change in size of a nodule in two time separated scans 23 Appendix C Examples of Image Viewer Configurations 1 The no java compatibility viewer 2 Single panel java viewer rots l lt B i gmt Cate Preview Mozilla Firefox File Ea View WHigtory Bookmarks Tools Help File Edit View Hitoy Bookmarks Tookt Help fag soe Series 0 Acquisition 1 Image 38 Compasbiiy gt ee SinglePanal _CT image Preview RetumHome Return to All Case Review Retur to Case Review Hep _ET image Preview RetumHome Retentie All Case Review Returnto Casa Rewiew Fap Study Date 1 Jan 2000 Bidii Biidii Sarex Arg 1 Zoom xi T Image Frame 19 of 43 Jangi POO 1S 281 F bape Far DA r Aden e FAm Window is 1500 level is 1600 Shee Thickness 00 mm semne Forward Reverse i iep Frame is 20 s Select Focus lindow is 1500 Level is 1600 al ae Quality 25 100 is 75 7 Pan Sing Lung Abdomen Bone Winedo
19. formation on how to access the system and the home page functions and Appendix B describers some of the intrinsic image analysis tools available in SIMBA Appendix C provides examples of some of the image viewing methods 1 1 SIMBA Applications The SIMBA system has been used for a variety of different applications all of these applications are supported by a single base SIMBA implementation Some of the main SIMBA applications are listed below 1 ELCAP multi center clinical studies The I ELCAP clinical study on lung cancer was a premier application of the SIMBA system and was developed in collaboration with the ELCAP research group at WCMC Further details of this application are given in section 2 2 Computer Vision Algorithm Research SIMBA has been the foundation system for computer algorithms research for the Vision and Image Analysis Group VIA in the School of Electrical Engineering at Cornell University The development of computer vision algorithms requires the collection of documented image data and the application of this data to both the training and evaluation of candidate algorithms All the computer research projects conducted by the VIA group employ the SIMBA system Further details on this application are provided in section 3 3 Image teaching system For this application users can view cases involving a sequences of images with documentation This has been used a medical teaching file in which users gain knowledge of case st
20. ge physical screen as shown in Figure 3 7 A stand alone java viewer that works in conjunction with SIMBA through a web link to SIMBA this may be configured to display similar to web methods 2 4 listed above This method provides for possibly higher performance when manipulating images 8 A stand alone java viewer that can display image files downloaded to the local computer this may be configured to display similar to web methods 2 4 listed above This is not linked to SIMBA but can operate on images downloaded from SIMBA Image viewing over the web has both advantages and disadvantages compared to a stand alone non web local computer solution Advantages include instant access to the latest data and documentation from a single archive the primary disadvantage is that it may take more time to display images if they have to be downloaded through the internet also running an application applet within the web browser has some resource and efficiency restrictions due to the browser host itself 3 3 Image Analysis There are two main components to the image analysis a individual processing of each case images and b that gathering of results from processed images across a set of cases For the former there are a number of image annotation and measuring capabilities that are already built into the system that are outlined in Appendix B Many projects are enhanced by the addition of custom designed analysis methods that are made available th
21. igure 1 All program system webpage 11 Select a program by clicking on the program icon For example clicking on Via Projects would take you to the program login page shown in Figure 2 Alternatively the URL may include the program qualifier and may take you to this page directly without showing the page in Figure 1 Da4vVision Management System File Edit View History Bookmarks Tools Help vi A i Ke eC ct lt E This program provides support for scientific image analysis projects Image and Data Management System select a Project and provide a Project key and a Personal key A Marking Platform Testbed YY Analysis Platform Testbed VD VisionGate 2D Segmentation VT VisionGate Experimental Segmentation TR Update review Testbed GA VisionGate V2 2D Segmentation GB VisionGate V2 3D Segmentation GC VisionGate V2 Truth Segmentation Project key Personal key RequestAccess Figure 2 Program qualified log on page In general you will be provided with the four items listed above before you can gain access to the system One exception to this is the public database accounts For these you do not need a project passkey and you can request directly from the login page a personal passkey which will be sent to you by email To access the system click on the name of the project enter the project passkey and your personal passkey and click on Request Access This completes the logon process and on successful logon you will b
22. including observer studies Clinical Studies Interactive teaching environment oo E Public image database This quickstart is directed at the first of these however the general format for all applications is the same For a Public image database you set up your own account and you may go directly to the URL to access the account for other account types you will need access information as outlined below 1 System Access Computer requirements for system use are a a network enabled computer with an up to date web browser and b for image viewing and annotation an up to date java plugin To log onto a specific project from a web browser you will need to know four items of information The URL of the system The name of the project The project passkey a Your personal passkey In addition if the URL does not specify the area of the application you will need to know the projects program An unqualified program only URL will provide you with a login screen similar to that shown in Figure 1 ne VIA Management System Mozilla Firefox schj lll a File Edit View History Bookmarks Tools Help VIA Image and Data Management System Please Select from the following programs p U l IC Acc Shotom Public access database open to anyone for access and downloading VIA P Ke ects Experimental Projects for registered users only E L A D Demonstration clinical accounts for registered users only F
23. ing For cases in which a standard report is required in the institutions local reporting system the system can generate an English language text report from that data in the data form that can be directly used for this purpose 7 Direct data download The system permits the direct data downloading of all data forms created by an institution This permits the institution to apply its own in house data analysis techniques to the data in the database 8 An SQL like database search It is expected that most users will download data and use their own analysis methods However the system does contain a built in search engine that can be used with an SQL like query syntax One advantage of this mechanism is that in addition to the data forms the search engine can also have access to the images and their documentation within the database 9 Anarchival messaging system The messaging system is like a global archival email that is viewable by all users It is particularly useful for informing users and documenting exceptional conditions in the study and also documenting their resolution Several of the above features provide for the maintenance of high protocol compliance and QA The protocol is built into the system and may be viewed interactively by the user Automated reports highlight any inconsistent data with respect to for example multiple radiologist readings and or delays with respect to the protocol Automated follow up lists are computed based on
24. lable Such forms may be deposited into a pending area which is not part of the main database Only when completed may they be submitted into the database 3 Integrated scheduling system The system contains all information for subject scheduling including the reason for scheduling obtained from data forms the contact information the study protocol and the history of attempts to schedule This information is integrated into a scheduling system that also provides work lists for the medical imaging device operators and an interactive follow up list to assist study coordinators in setting subject appointments 4 Security system with multiple user classes SIMBA has a special security system that provides access models for a number of different user classes For example a class of visiting radiologist may be established which provides such users with a limited set of de identified images via an active work list Another account class for example is designed for QA monitoring by center personnel 5 Reports for all main system activities of the project are available to authorized users Currently 16 different reports may be generated for different aspects of the projects operation A number of these reports that report discrepancies or inconsistencies are active in that they provide direct links to data forms at issue and assist the user 6 Radiological reporting system The CT radiology data forms provide all information for a standard chest CT read
25. lann otation1 256 256 E Ce bBX0037 133211 00 02 Pa lssz7_ las lso Pew J Figure 2 Example Case Review This view presents thumbnail images of all the image data sets associated with the case and also lists all the annotations that have been marked on those images For each image options are provided to view the full image data or to view the documentation for that image 3 2 Image Viewing Image viewing is typically requires significant computer resources both in terms of memory and processing Furthermore different projects involve different amounts of image data for example in some projects such as face recognition the viewing of small two dimensional images 10 KB may be simply viewed while in others large 3D images or movies 1 GB may need to be manipulated and annotated SIMBA is designed to work with a wide variety of client computers some of which may have very limited resources and some of which may have extensive resources specifically for managing images To support this range of client resources and range of applications SIMBA provides at total of eight different viewing modes that are user selectable two of which use stand alone versions of the java viewer Most applications and computers are adequately served by the usual default mode number 3 in the list below See appendix C for some screen shots of the basic web based options the main options are listed below in order of increasing resource needs 1 View images o
26. n a client without a java viewer This is an emergency mode for clients that do not have a java plugin installed or have an old or damaged java plugin Very few computers today do not have a working java plugin This facility permits the viewing of most of the image types it does not support either the creation or display of image annotations 2 Embedded Java viewer with a single panel This may be used on computers with a small screen resolution 800 x 600 Annotations and other options are accessed through a pop up window 3 Embedded java viewer with annotations panel This is similar to 2 except that the the annotations and related images are always directly available but requires a wider screen minimum 1024 pixels wide This is the default viewing mode for many applications 4 Two panel viewer for image comparisons This mode permits two images to be displayed and compared side by side Additional functions and annotation access are through a popup window 5 Two panel viewer for across case comparisons In this mode there are two virtual screens overlaying each other One contains a two panel viewer and the other is the SIMBA web system The user selects which of these two screens to view at a given time Images selected in SIMBA may be viewed in the viewer screen In this 6 Two panel viewer for large screens screen size larger than 1800 pixels wide This is simiar to 5 above except that both screens are displayed simultaneously on the lar
27. nalysis when the they have been added by project customization 5 Detailed Control Panel This panel located to the right of the main menu provides access to the main system features 6 Case Selection is accomplished from the control panel There are several search and selection functions that facilitate the access to both case subsets and individual cases 7 Project Options this section of the control panel provides buttons to optional customized features of the project 8 Password management provides for changing your personal passkey and to logging off the system 9 System help This menu provides links to help information for some of the main system activities 10 Boundary Marking Help At this time there is no other user s manual for the system beyond this quickstart document All user information for the main system functions and for a specific project is accessible on line from the project home page In addition further information on specific activities is often provided by yellow help buttons located on that activities page 14 Appendix B SIMBA System Standard Image Analysis Tools In this appendix additional details of our standard analysis tools are presented with examples These tools include a number of methods for measuring image regions such as lesions and performing a size change analysis These tools may be used for example for the direct growth analysis of lesions on CT image data sets or for any quantitativ
28. ncludes any additional image documentation The non DICOM images may be annotated and viewed using the same tools as for the DICOM images 3 An Interactive java image viewer is used to view image data sets through the web using secure encrypted data transfer SSL The viewer is a very important component of the web interface It provides conventional Radiologist viewing options windowing zoom pan etc and works on CT mage data sets of all sizes In addition it provides additional viewing modes such as nodule montage side by side image comparison and computer assisted boundary marking tools for the interactive outlining of lesions It can manage both DICOM and non DICOM image formats 4 Image data download in different data formats DICOM image data may be downloaded through the web in several different data formats 2 3 Image Analysis Tools In our ongoing research we have pioneered the use of automated volumetric methods for establishing pulmonary nodule growth rates We have made these methods available to the clinical studies through a simple web based interface In addition we have developed a number of different general purpose interactive measuring tools that permit the comparative viewing and analysis of time separated image sets These tools may be used for example for the direct growth analysis of lesions on CT image data sets or for any quantitative image analysis function on any image modality The currently available image anal
29. or instructions on how to do a Select Focus Window 1500 No Measur Show Re eer er arking Zoom in RAN in Orie in a 2 Click on the Hide button next to the Reference Bnd to remove any reference markings lt Cave a ss ae 3 Click on Save As button to record your boundary marking Single Presets Lung Open n Settings Figure 3 Side by side large screen presentation with the two panel viewer on the left and the SIMBA web screen on the right Customized image analysis for a single image is achieved by clicking on the image thumbnail in the case review shown in Figure 2 When available clicking on the image thumbnail will take you to a project specific page that facilitates the setting of parameters and the image processing functions If specified the derived images from the image analysis will subsequently be shown as thumbnails in the case review and they may also be selected by these thumbnails for further processing Web pages that provide project specific functions for analysis across cases and information that describes how to use the customized project package are accessed from menu items on the project home page see Figure 3 in Appendix A 10 Appendix A SIMBA Image Management and Analysis System Quickstart The image management and analysis system provides support for a variety of web based image applications Applications are categorized into four broad areas Research projects
30. ot from the emphysema web page is shown in Figure 14 Click ona nodule candidate for further inspection Figure 13 Snapshot of the lung nodule detection tool Nodule candidates generated by the system are displayed on a coronal lung view computed from the CT axial image slices Clicking on a nodule candidate permits the user to select and apply one of the nodule measuring tools 22 Scan 1 Scan Date Thu Jan 6 11 26 17 2000 Emphysema Index 4 63 Lung Volume 6012 ml Coronal Lung Image Emphysema Index Image Emphysema Index Key None Severe Figure 14 Snapshot of an emphysema analysis The Emphysema and lung volumes are reported in addition a coronal visualization of the lungs obtained from the CT scans is shown on the left and the distribution of emphysema within the lungs is shown on the right Two Panel Web based viewer To facilitate comparison of images we have developed a viewing system that allows time separated images to be displayed side by side on the web In this way there can be uniformity in the way that images are marked when assessing for change Each of the panels has the full functionality of the single panel viewer and allows for automated 3D analysis as well as manual drawings and editing A web snapshot of the two panel viewer is given in Figure 15 aoa PROD ay E T image Previews if Feehan Hore Retum fo Dase Fisiere E FROG PRODI Series 2 Ate 1 l Link FROG PROD Series 2 Aco 1 S
31. rough the standard SIMBA API J VIA image View Split Mode 2 M Ble ECR View History geokmaks Took Heb lt gt C OR tr fledde via_cornell ecujcgrbinidataciuprets cg E G Bx0009 8x000 Series GAcq 1 Zoomx1 0 80015 80015 Zoomd 0 Jan 01 2000 13 25 17 image 26 43 082mmxO82nmx7 Sm Oct 31 1990 11 40 26 00 o7omms07 smmxsomm BX Cornell University Blind Database Home Case Review and Data Entry image Management System a Locate by View All Cases Case ID User Preferences Description Help for Main Activities Record No Region and Change Analysis View all Cases DICOM Image Download 7 Image Upload Case Management Messages View Messages Creata New Message E Vision and Image Analysis Group General Information and Messaging Management System Updates Image Viewing elect a case either click on View AN C enter tr i Cases then click o iter the Case number in the Case ID field and then click on d utton on the image List button for the desired the associated Go bt ew Dutton below the image of interest 3 Click on the LZ button in the image viewer for assistance on using the image viewer Image loaded with bookmark Change Window 4 Click on Return Home When done lt Navigation gt Windowsing Measure Boundary s Navigation gt Windowing Boundary Marking lt gt Change Undo v lt m iaa pA j to view as outlined above Click on the button f
32. sed in clinical studies that are audited by the FDA 1 2 SIMBA foundation capabilities The generic SIMBA system provides a facility for managing image based projects Each project has the following set of standard features 1 Secure user access A double password entry system is used Users are assigned to one of a number of different classes having different access permissions The administrator class has the most permissions and is permitted the management of other users The research class has the minimum set of permissions and only allows viewing some of the data Case management data is organized into cases in the clinical setting a case would refer to an individual subject Image data sets and other study documentation are managed within cases Image management For each case a set of images may be managed Images may be uploaded to the system through a web interface or may be received directly from say a PACS system using a DICOM protocol or both There are two image archives a general image archive that accepts many image data formats including DICOM and a DICOM only image archive that also manages the DICOM header information All images in the system may be interactively viewed through the web browser using a java applet or downloaded for local viewing All images may be annotated with specific views or with boundary marks The images may be analyzed by the inclusion of project specific functions Image Documentation All upload
33. the most recent information in the system to ensure that subjects are recalled on time 2 2 On line Image Archive The image archive system performs similar functions to a conventional PACS system in that it provides a repository of medical image data sets in DICOM format It also has the ability to manage images such as pathology slide images which are in different image data formats In addition to the integration of this archive with the other components of the EMS the features of the image archive that extend beyond standard PACS capabilities are as follows 1 Direct receipt of data from hospital RIS A special relay computer has been developed that transfers DICOM image studies from a hospital RIS to the central image archive The relay may de identify the data and encrypts the data for internet transfer using the SHH protocol This greatly simplifies the collection of medical images from different sites The technologist operating the imaging device simply pushes the images for a study to one additional DICOM device the local SIMBA relay 2 Management of a diverse set of image data formats and an image upload capability In addition to DICOM formatted data CT CR MR PET etc the system can also accept other image data such as microscope images and analysis images Such images may be interactively uploaded from a web enabled digitizing microscope or for any conventional computer system Uploaded images have an associated data form which i
34. udies by reviewing documented case sequences There is also an author facility that permits author users the ability to upload images organize them for presentation and document them Public Image database The development of computer algorithms for image analysis requires the availability of a database of example images with annotations that provide correct analysis outcomes This database is then used for both training and evaluating the performance of the computer algorithms Two public documented image databases have currently been created with SIMBA Observer study image marking experiments The performance of experts in marking images specifically radiologists marking the boundaries of lesions in medical images is obtained by observer studies in which a set of markers annotate an evaluation set of images in under a controlled protocol SIMBA supports observer studies through the web browser so that expert markers may be geographically distributed and do not have to do the study at the same time Image measurements for clinical trials Recently a number of drug trails have used lesion size measurements over time as an indication of response to therapy The SIMBA clinical marking application provides for the documented uploading of images the assignment of these images to radiologists to measure over the internet and the recording the measurements made by the radiologists This part of the system conforms to 21 CFR part 11 and therefore may be u
35. wing Head Nex View Change Use Java Viewer Ho Lear H3 Grd Boundary Undo Show rat 3 The java viewer with annotations bose Freee Moly rotor a Eile Edit View Hrjtory Bookmarks Jocis Help BX0009 Single Annotation CT image Preview Retum Home Retan to All Case Review Return to Case Review me E0009 BXD009 Series 0 Acq 1 Zoom t 0 Anmotations Browse Cases Jan 01 2000 132517 Image 24 43 082mm 0 82men x 7 Smee Li tr Annotahon titleicommeats Load Load Nent Reference Bnd Hide e Show Vertces Edit Commands Edit Nudge Undo Show Vertces Edit Boundary Hide f 4 Annotation Commands New Ect Save as Setect Focus Window 71500 No Measur ShowRet al Zoom in Level 440 No Grid Pan Singe Preses Lung gt Settings 24
36. ysis capabilities are listed below further details and examples of these tools are presented in Appendix B 1 Fully automated three dimensional volumetric nodule growth analysis A user requests a nodule analysis by simply clicking with the mouse on the image of a nodule displayed on a web browser with the java viewing tool outlined in the previous section The automated system then determines the volume of the nodule In the case of sub solid nodules the system computes a volume for both the solid and sub solid nodule components Once volumes for a nodule from multiple scans are available a growth analysis can be performed The system reports on growth rates obtained by the size change for all possible pairs of time separated scans of a nodule 2 Manual marked area lesion growth analysis In many cases of pulmonary nodules the totally automated methods may produce the most accurate results However for other lesions with less well defined margins and for cases where the scanner parameters are significantly different automated methods cannot currently be used For these situations interaction between the radiologist and the system is necessary For the area growth analysis method the radiologist selects the best central image of a lesion and carefully marks its boundary Growth analysis is performed similar to the automated method but is based on that area enclosed by the boundary 3 Manual marked volume lesion growth analysis The manual volume
37. zilla Firefox File Edit View History Bookmarks Tools Help Selected Image Region Figure 5 Side by side comparison of the areas marked by the user presented for review The regions may then be readjusted by the user to ensure that they are consistent The same growth analysis may be applied if standard caliper measures are applied to a representative two dimensional image the only difference is in the visualization of the measure A RECIST uni dimensional measure is represented by a single line as shown in Figure 6 And a WHO perpendicular lesaion measure is visualized as shown in Figure 7 Growth Analysis Mozilla Firefox File Edit View History Bookmarks Tools Help Selected Image Region Figure 6 growth measurement using a uni dimensional RECIST method Growth Analysis Mozilla Firefox File Edit View History Bookmarks Tools Help Selected Image Region Figure 7 growth measurement using the bi dimensional WHO measure 18 3 Manual marked volume lesion growth analysis The manual volume method is similar to that of the manual area method with the difference being that each of the 2D measurements comprising the volume of the lesion is included The boundary is determined on each image that the lesion is visible and a composite volumetric rather than area based growth estimate is performed This method has been adopted by the LIDC as the basis for documenting pulmonary nodules in this public reference image database ref
Download Pdf Manuals
Related Search