CTSC:TTIC.progress

Back to Collaboration:Harvard_CTSC

Image data analysis and visualization tools

3D Slicer technologies particularly appropriate for dissemination to the clinical community were identified by CTSC Imaging Core investigators (Ron Kikinis, Clare Tempany, Randy Gollub, Jeffrey Yap, Wendy Plesniak). A new Slicer software module for computing the Standardized Uptake Value in PET imaging, and tools to visualize PET/CT studies were also specifically designed and developed. A stable software binary was built and tested with extensive expertise and support from Kathryn Hayes, so that this module could be made rapidly accessible to the broader clinical translational community.

Two tutorials have been developed and tested for the CTSC community demonstrating how to use this version of 3D Slicer to: a) perform standard RECIST analysis to assess tumor response to therapy in MRI Volumetric studies, and b) perform SUV analysis to assess tumor response to therapy in PET/CT studies. (Delivered on 11/23/09, and delivered to RSNA 2009 on 12/01/09).

We also continue ongoing work to build 3D Slicer’s functionality, focusing on delivering cutting edge analysis techniques and algorithms to clinician investigators interested in experimenting with them. Currently, an implementation of DCE-MRI analysis using the Tofts Kinetic Model to assess tumor response to therapy in perfusion studies is being developed by Junichi Tokuda, and is being modified to be more user-friendly and extended to function across all platforms by Wendy Plesniak.

3D Slicer’s Extension Manager is another newly developed mechanism for allowing external software modules to be compiled into and disseminated with the 3D Slicer platform. This mechanism allows algorithm developers to more rapidly introduce new functionality into the Slicer platform, and permits individual clinician scientists to “tailor” their individual Slicer build to include only the functionality that supports their workflows. Kathryn Hayes continues to work with numerous investigators, helping them to integrate their new developments (e.g. algorithms for segmentation, skull stripping, cortical thickness measurement) into 3D Slicer.

Data management tools

An set of initial XNAT Enterprise use-case assessments were conducted in the biomedical research community to determine if their local data handling needs could be appropriately handled by XNAT’s technical capabilities, and if so, to determine how best to support their workflows. This work was conducted by Wendy Plesniak, along with Yong Gao and Mark Anderson. Four research groups with large retrospective studies to manage were identified: Ellen Grant’s and Rudolph Piennar’s laboratory at Children’s Hospital Boston; Brad Dickerson’s laboratory at Massachusetts General Hospital, Simon Warfield’s Laboratory at Children’s Hospital Boston, and several projects within the National Center for Image Guided Therapy (NCIGT) (PI Ferenc Jolesz) at Brigham and Women’s Hospital. Interviews were conducted with each of these groups, requirements for data management, and descriptions of their workflow were captured in a report on our wiki to guide our approach: (links for each group can be found here: [1]).

To develop our understanding of XNAT, three available approaches for anonymizing, bulk uploading and applying metadata to large retrospective studies were explored. These three approaches included a) using XNAT-provided desktop GUI-based tools, b) using XNAT-provided batch scripts, and c) using XNAT’s REST-ful web services interface. Each approach requires a significantly different workflow, has different benefits and limitations, and offers a different user experience. Two were found to be most appropriate for our selected use-cases. The command line scripts, which have been used to store data for the Dickerson and Grant/Piennar lab, have also been used for four projects within NCIGT: Functional Data for Neurosurgical Planning; NCIGT Intra-operative MRT Glioma Resection; NCIGT-Harvard-BWH Neurosurgical Intraoperative Image Database GENESIS format data; and the Prostate MRI Database. These uploads have been tested using XNAT instances at CHB, at central.xnat.org, and most recently at BWH (described below). Finally, these original XNAT command-line scripts were consolidated into a single master script by Yong Gao to simplify the bulk upload of large data collections. This consolidated script is currently being tested on data from the Warfield lab. Additionally a set of scripts designed to use XNAT’s web services interface were highly tailored to the individual use case for upload and query, and these are being tested in the Grant/Piennar lab. This work has also been the collaborative focus of Anderson, Plesniak and Gao, working closely with investigators from each use-case.

In the interest of integrating XNAT Enterprise and 3D Slicer, Wendy Plesniak worked with NA-MIC colleague Curtis Lisle to plan the architecture and user experience for Slicer’s integration with XNAT Enterprise REST-ful web services. (Plesniak has already developed the Fetch Medical Informatics (FetchMI) module that interfaces Slicer to an extensible suite of web services.)

An installation of XNAT server was recently set up for use within the Surgical Planning Laboratory at Brigham and Women’s by Mark Anderson. In order to install this instance, Java, Tomcat and psql were installed, and the server was configured to store DICOM data (similar to the configuration at central.xnat.org) on host gridftp.bwh.harvard.edu. This instance and is currently being tested (with some of the data from the NCIGT Prostate MRI Database.)