NIH: 2006 All NCBC Meeting

From NAMIC Wiki
Jump to: navigation, search
Home < NIH: 2006 All NCBC Meeting

Logistics

All information about the meeting is available at the NCBC AHM 2006 Website.

Dates: Mon-Wed July 17-19 at NIH.

Confirmed NA-MIC Delegation: Please register here: http://pub.nigms.nih.gov/generic_meeting/registration.cfm?id=31

  • Ron Kikinis, Sun-Weds afternoon
  • Bill Lorensen, Sun-Tue evening
  • Steve Pieper, Sun-Wed evening
  • Will Schroeder, Sun-Wednesday noon
  • Steve Wong, Sun-Wed afternoon
  • Sonia Pujol, Sun-Thursday
  • Ross Whitaker, Sun-Tuesday afternoon
  • Marty Shenton, Sun-Wed

Some points to note:

  • There is a meeting hosted by Ron on Monday, July 17th, 5:30-7pm of the NA-MIC PIs in room 2As.10 in Natcher (reserved by John Whitmarsh's assistant Jennifer Villani)
  • Assignments of folks to the breakout sessions:
    • Yellow Pages - Bill, Will
    • Ontologies - Steve Pieper, Ross
    • System Biology - Steve Wong, Marty Shenton
  • To bring your personal laptop into the meeting, please print, fill out, and bring this Property Pass with you to the Lister Hill Center on NIH campus.

Accommodation

Hyatt Regency Bethesda
7400 Wisconsin Ave
Bethesda, Maryland 20814

Tel: 301-657-1234
Fax: 301-657-6453

Per Peter Lyster, please call 800-233-1234 to make hotel reservations by Friday, June 16.
Refer to the Function Name: "ROAD".

Agenda

Rons talk: Monday, July 17, 9-9.30am: NA-MIC kit

Abstracts and Posters

Poster Instructions

Hyatt Regency Bethesda, One Bethesda Metro Center (7400 Wisconsin Ave), Bethesda, MD 20814

Tracking information

  • Structural Shape Analysis: Fedex Tracking with 8583 8301 6218

Submitted Abstract Files


NA-MIC Kit

Will Schroeder, Bill Lorensen, Andy Cedilnik, Jim Miller, Sebastien Barre, Dan Blezek

Abstract Text

Medical image computing researchers often face the problem of moving promising algorithms from inception to clinical application. Algorithm developers lack the time and resources to engineer their code for robustness and compatibility, while end-users are anxious to try new techniques but require well designed and tested user interfaces to make practical use of them. The NA-MIC Kit is a collection of software and methodology specifically designed to address these problems and facilitate the rapid advancement of the field. It consists of three major types of software technology: programming toolkits (e.g., VTK and ITK), end-user application software (e.g., Slicer), and system infrastructure (e.g., LONI, CMakem CPack, DART2). In addition, the NA-MIC Kit addresses issues of usability, software process including quality assurance, community building and licensing. These technologies are integrated in a consistent framework that facilities the transition of ideas to usable, quality software implementations. Besides showing preliminary results benefitting the medical image computing community, components of the NAMIC Kit have been successfully adopted by other large projects such as KDE, one of the world's largest open source software systems (KDE is the Linux windows environment).

Keywords: VTK, ITK, Slicer, LONI, CMake, DART, CPack, KWWidgets, quality software process

Poster PPT and PDF

Poster PDF Poster PPT

fMRI Analysis

Title: SPATIAL REGULARIZATION FOR fMRI DETECTION

Ron Kikinis, Wanmei Ou, Polina Golland, William Wells III, Carsten Richter, Haiying Liu, Wendy Plesniak

WO, PG, WW: MIT Computer Science and Artificial Intelligence Laboratory, Cambridge MA

RK, WW, CR, HL, WP: Surgical Planning Laboratory, Brigham and Women’s Hospital, Boston MA

Abstract Text

We have developed a method for brain activation detection that employs Markov Random Fields (MRF) as spatial smoothing priors necessary due to the low signal-to-noise ratio of the fMRI signal. Furthermore, we extended the MRF prior to include anatomical information. The anatomical prior, in the form of a segmented MRI scan, biases the activation detection towards the gray matter and inhibits smoothing of the activation maps across tissue boundaries. We have validated the method on a set of digital phantoms and a set of fMRI scans.

We are currently translating this work into 3D Slicer, an open source medical image analysis and visualization package developed and supported by Core2 activity, and are exposing the functionality through Slicer’s fMRIEngine module. The fMRIEngine module is designed to provide a framework for an extensible suite of activation detection algorithms; it currently provides a General Linear Model-based approach with which the MRF and anatomical priors can be used. Algorithms have been implemented in C++ as extensions to Kitware’s Visualization Toolkit (VTK) classes. A graphical user interface has been developed which allows a user to select a parametric map of activation from within the fMRI analysis workflow, to (optionally) load a segmented MRI scan and configure anatomical prior probabilities, to choose and configure a method for density estimation, and to configure the Meanfield approximation. Resulting color-coded posterior probabilities of brain activation can be viewed in Slicer’s 3D viewer and compared to original parametric brain activation map.

This prototype has been tested on the same datasets as used to validate the algorithms originally and has produced identical results. It is currently available to the wider community in the Slicer2.6 CVS repository but is not yet distributed with the Slicer2.6 release as refinements to the embodiment are ongoing. Our plans include incorporating this processing into Slicer tutorials provided to the user community, releasing the software with Slicer2.7, moving the implementation from VTK to ITK, and releasing the code into the ITK library to make the method available to a broader community.

Publications:

Wanmei Ou, Polina Golland. From Spatial Regularization to Anatomical Priors in fMRI Analysis. In Proc. IPMI 2005: Information Processing in Medical Imaging, LNCS 3565:88-100, 2005.

Wanmei Ou. fMRI Detection with Spatial Regularization. MIT Master Thesis, May 2005.

Poster PPT and PDF

Poster PDF Poster PPT

DTI Analysis

(Martha Shenton, Marek Kubicki, Sylvain Bouix, Marc Niethammer, Lauren O’Donnell, Carl-Fredrik Westin, Tom Fletcher, Martin Styner, Ross Whitaker, Guido Gerig, Ron Kikinis)

Abstract Text

Neuroimaging studies over the last two decades have led to major progress in delineating gray matter abnormalities in schizophrenia. By comparison, far less is known about white matter abnormalities in schizophrenia, especially those affecting white matter fiber tracts that connect the frontal and temporal lobes, tracts that have long been thought to be abnormal in this disorder. With the development of diffusion tensor imaging (DTI), however, we are now able to investigate more fruitfully white matter abnormalities in schizophrenia. As part of the NA-MIC project we have focused on white matter fronto-temporal connections, including the uncinate fasciculus, arcuate fasciculus, and cingulum bundle. One example of this work is our collaboration with computer scientists at Brigham and Women’s Hospital and MIT, where we have developed diffusion tensor based anisotropy measures to characterize white matter abnormalities in schizophrenia. Here, we present findings from this collaboration where we have applied anisotropy measures to quantify diffusion properties of fronto-temporal fiber bundles (cingulum bundle, fornix, uncinate fasciculus and inferior occipito-frontal fasciculus). Additionally, we note that an important challenge in DTI is to be able to automatically segment white matter into known fiber bundles. Fiber tractography has begun to fill this niche and is a central tool for tracing fibers over the entire brain. However, the resulting fibers still need to be organized in an anatomically meaningful way. Accordingly, and in collaboration with computer scientists at MIT, we have developed a new algorithm to automatically segment fibers into organized and meaningful fiber bundles (O’Donnell). Moreover, we note that quantifying properties of a bundle by analyzing individual fibers can be cumbersome. To address this issue, and in a team effort with collaborators from Brigham and Women’s Hospital, we have developed a novel method to provide a continuous representation of each bundle. This new representation greatly facilitates the computation of geometric and diffusion properties over an entire fiber bundle. We are currently in the process of applying these new techniques to study white matter properties in schizophrenia.

Poster PPT and PDF

Poster PDF Poster PPT

Structural Analysis

Title: STASTICAL SHAPE ANALYSIS OF BRAIN STRUCTURES

Author: Styner Martin, Oguz Ipek, Gerig Guido, Nain Delphine, Tannenbaum Allen, Golland Polina, Bouix Sylvain, Niethammer Marc, Fletcher Tom, Whitaker Russ

SM, OI, GG: University of North Carolina, Chapel Hill, NC
ND, AT: Georgia Tech, Atlanta, GA
GP: MIT, Cambridge, MA
BS, NM: BWH, Boston, MA
TF, RW: University of Utah, Salt Lake City, UT

Presenting author?
Keywords: Shape Analysis, Brain Morphometry, Neuroimaging

Abstract Text

Quantitative morphologic assessment of brain structures is routinely based on volumetric measurements. Volume changes are intuitive features as they might explain atrophy or dilation due to illness. On the other hand, structural changes at specific locations are not sufficiently reflected in volume measurements. Shape analysis has thus become of increasing relevance to the neuroimaging community due to its potential to precisely locate morphological changes between healthy and pathological structures.

Within the NA-MIC network, the following key problems in structural shape analysis are being tackled in a joint effort: shape representation, correspondence, statistical testing, and tool development. A first, basic framework and set of tools for statistical shape analysis based on a sampled local spherical harmonic shape representation (SPHARM-PDM) has been implemented. Within this shape analysis framework further methodological developments have been developed:
Representation: Two additional shape representations complement SPHARM-PDM. A novel spherical wavelet based representation allows a hierarchical decomposition and an increased sensitivity of the statistical analysis through a reduction of the number of shape features. Another novel particle-system based representation allows the analysis of structures of non-spherical topology, such as the cranium, as well as is insensitive to noise-induced topological variations.
Correspondence: A new method is in development for the correspondence of surface points across different datasets using a population-wise optimization of a curvature metric. Based on our earlier comparisons studies, this method will enhance the anatomical correctness, the statistical modeling properties, as well as the statistical sensitivity.
Statistical Testing: We developed methods for the computation of group discrimination as well as non-parametric statistical hypothesis testing. Also, correction for multiple-comparison testing has been incorporated into the statistical shape analysis framework based on two complementing approaches, family-wise error control via permutation testing and false discovery rate control. These correction methods are crucial in enhancing the sensitivity and the power of the analysis.
Tool Development: Based on closed interaction with the engineering core, shape analysis tools were developed within the Insight Toolkit (ITK) framework. These tools have been incorporated into the LONI pipeline software, and disseminated to several clinical collaborators within and outside the NA-MIC network.

We applied the SPHARM-PDM shape analysis tools on data from a male schizo-typal personality disorder (SPD) study (VA Brockton, BWH) available within the NA-MIC data repository. The results show significant shape differences located in the posterior and anterior parts of the caudate head on both hemispheric structures, but enhanced on the right caudate. Additional results computed using the same tools on a new female SPD study showed similar findings with enhanced sensitivity due to an enlarged sample size. Both of these datasets have also been analyzed using the spherical wavelet based representation based on the SPHARM correspondence. These results both confirmed the earlier ones, as well as extended the significance pattern in regard to a hierarchical decomposition. Current studies are in progress using the novel particle based approach, as well as using the enhanced curvature based correspondence method.

Poster PPT and PDF

Poster PDF Poster PPT

Dashboards: A brief history of Dart

Dan Blezek, Jim Miller, Bill Lorensen, Andy Cedilnik

DB, JM, BL: GE AC: Kitware

Abstract Text

For any large software project to be successful in the long term, it is imperative to develop a testing infrastructure to monitor the quality of the system as it develops and matures. In general, developers of such systems are averse to this requirement. As a result, code begins to become stale, and developers no longer understand the ramifications of their changes, and are unable to refactor code from the past. So they press on in the hopes that their foundational code does not crumble beneath them.

NAMIC seeks the higher ground. The following is an account of the history of our software quality system called Dart.

In 1997, General Electric added a new quality initiative, called Six Sigma. At GE Research, we focused a collection of our quality projects on the development of one of our software toolkits, the Visualization Toolkit or VTK (http://www.vtk.org/). We integrated our original 14 projects into an automated system that collected their outputs and integrated them into an online, daily VTK dashboard.

In 1999, the National Library of Medicine commissioned the development of an open source, cross platform project called the Insight Segmentation and Registration Toolkit, ITK (http://www.itk.org/). As part of GE Research's contribution to the ITK effort, GE Research developed the first version of Dart. Dart's goals were

  • Remove the dependence on tcsh, awk, sed and cron scripts to perform a build and test sequence
  • Allow testing machines from around the world to submit test results to a Dart server
  • Separate the data from its presentation
  • Apply the Dart testing system to a variety of project (ITK, VTK, VXL)
  • Make the testing system itself open source

Dart met its original design goals and was successfully applied to many software projects. Dart clients were easy to use and allowed for testing machines to be distributed around the world. The Dart server allowed anyone to view the results of a test sequence and monitor the software development process. Dart allowed a cross-platform system to be tested in multiple configurations, though the server side of Dart was still difficult to maintain.

In 2004, NIH sponsored the National Alliance of Medical Image Computing, NA-MIC (http://www.na-mic.org/) as part of NIH Roadmap for Medical Research, Grant U54 EB005149. GE Research is developing the next generation of Dart as part of NA-MIC. The goals remain broadly the same, however, two new goals have been identified

  • Simplify the Dart server setup and maintenance
  • Allow for longitudinal or temporal analysis of test results

To this end, we introduce the new version of Dart. We affectionately refer to the previous version of Dart as Dart Classic. The new Dart still accepts build/test results in the Dart Classic format. Dart has been completely rewritten in Java. It uses an embedded web server and servlet engine (Jetty) and an embedded database (Derby). XML-RPC is used to transmit build/test results to the Dart server. Dart is distributed as two jar files. The first jar file, DartServer.jar, contains everything to create and run a Dart server managing several Dart projects. The second jar file, DartClient.jar, is a small utility to shutdown a server, refresh its resources, query its status, and can be used as an XML-RPC messenger.

Dart has recently passed the 1.0 release milestone and is moving into scale-up testing and debugging for broad adoption across the multitude of NAMIC software projects. More information may be found on the Dart homepage: http://na-mic.org/Wiki/index.php/Dart2Summary