2009 Annual Scientific Report

Back to 2009_Progress_Report

Guidelines for preparation

• 2009_Progress_Report#Scientific Report Timeline - Main point is that May 15 is the date by which all sections below need to be completed. No extensions are possible.
• DBPs - If there is work outside of the roadmap projects that you would like to report, you are welcome to create a separate section for it under "Other".
• The outline for this report is similar to the 2008 and 2007 reports, which are provided here for reference: 2008_Annual_Scientific_Report, 2007_Annual_Scientific_Report.
• In preparing summaries for each of the 8 topics in this report, please leverage the detailed pages for projects provided here: NA-MIC_Internal_Collaborations.
• Publications will be mined from the SPL publications database. All core PIs need to ensure that all NA-MIC publications are in the publications database by May 15.

Introduction (Tannenbaum)

The National Alliance for Medical Imaging Computing (NA-MIC) is now completing its fifth year. The Center is comprised of a multi-institutional, interdisciplinary team of computer scientists, software engineers, and medical investigators who have come together to develop and apply computational tools for the analysis and visualization of medical imaging data. A further purpose of the Center is to provide infrastructure and environmental support for the development of computational algorithms and open source technologies and then to oversee the training and dissemination of these tools to the medical research community. We are currently in year two of our second set of Driving Biological Projects (DBP), three of which involve diseases of the brain: (a) brain lesion analysis in neuropschiatric systemic lupus erythematosus; (b) a study of cortical thickness for autism; and (c) stochastic tractography for velocardiofacial syndrome (VCFS). The fourth DBP takes the Center in a very new direction, (d) the prostate: brachytherapy needle positioning robot integration.

Over the past five years, NA-MIC has made substantial progress toward the attainment of its major objectives. In year one, the Center focused on forging alliances amongst its various cores and constituent groups to assure that the efforts of the cores were well integrated toward the attainment of common and specific goals. To that end a great deal of effort went into defining the kinds of tools that would be needed for specific imaging applications. The second year emphasized the identification of key research thrusts that cut across all cores and were driven by the needs and requirements of the DBPs. This led to the formulation of the Center's four main technical themes: Diffusion Tensor Analysis, Structural Analysis, Functional MRI Analysis, and the integration of newly developed tools into the NA-MIC Tool Kit. The third year of Center activity was devoted to the continuation of collaborative work to develop solutions for the various brain-oriented DBPs. The fourth year was focused on translating collaborative knowledge and work to a new set of DBPs. In the current fifth year, a number of projects have made sufficient progress to warrant introduction as modules in Slicer, thereby making the Core 1 algorithms available to the general medical imaging community. Some of these algorithms are quite general and can be used for purposes far broader than the original DBPs. For example, a new point cloud registration algorithm developed for the prostate brachytherapy needle positioning project also can be used for DWI registration. Likewise, work on DTI/DWI tractography has been applied to the segmentation of blood vessels and soft plaque detection in the coronary arteries.

Year five progress with respect to the current DBPs is relevant to the scope of this Annual Progress Report. As mentioned, we currently have three projects in the area of neuropsychiatric disorders: Systemic Lupus Erythematosis (MIND Institute, University of New Mexico), Velocardiofacial Syndrome (Harvard), and Autism (University of North Carolina, Chapel Hill). A fourth project from Johns Hopkins and Queens Universities involves the application of core technologies to imaging/robotics-guided treatments in prostate cancer. A number of papers have been published that specifically acknowledge the NA-MIC, and significant software development is continuing as well.

Section 3 outlines specific aims fulfilled this year by the four roadmap projects: Section 3.1 is devoted to the Stochastic Tractography Approach for Velocardiofacial Syndrome; in Section 3.2 we describe the application of our work to Brachytherapy Needle Positioning for the Prostate; Section 3.3 outlines the Brain Lesion Analysis in Neuropsychiatric Systemic Lupus Erythematosus project; and Section 3.4 describes the Cortical Thickness for Autism project. For all of these projects, a synergism of effort has produced working computer modules that are user friendly and accessible to both medical researchers and clinicians.

Section 4 describes year five work on the four infrastructure topics. These include: Diffusion Image analysis (Section 4.1), Structural analysis (Section 4.2), Functional MRI analysis (Section 4.3), and the NA-MIC Toolkit (Section 4.4). Many of the algorithms produced by Cores 1-3 have been inegrated into ITK and Slicer, including those concerning shape analysis (e.g., spherical wavelets), new segmentation algorithms (for DTI/DWI tractography and the segmentation of the prostate), and new approaches to registration (e.g., based on particle filtering).

Finally, we use the last three sections of this Annual Progress Report to highlight some of the work that the Scientific Leaders believes is particularly significant to the the overall goals of the Center. Section 5 summarizes the benefits of several advanced algorithms, gives a description of the growing NAMIC-Toolkit, and documents the scope of our efforts in technology transfer and outreach. It is essential to emphasize that although the algorithms emanating from this Center were developed to solve specific clinical problems raised by the DBPs, in application, most of these algorithms have far more general utility and far greater potential impact on the medical imaging technical base. Section 6 draws attention to the impact and value of our work on biocomputing imaging at three different levels: within the center, within the NIH-funded research community, and externally in the national and international community. Section 7 provides some updated timelines with specific milestones achieved by the various NA-MIC cores. Section 8 lists publications that acknowledge NA-MIC support, and to Section 9, we append the External Advisory Report and our considered response.

Clinical Roadmap Projects

Roadmap Project: Stochastic Tractography for VCFS (Kubicki)

Overview (Kubicki)

The goal of this project is to create an end-to-end application that would be useful in evaluating anatomical connectivity between segmented cortical regions of the brain. The ultimate goal of our program is to understand anatomical connectivity similarities and differences between genetically related schizophrenia and velocardio-facial syndrome. Thus we plan to use the "stochastic tractography" tool for the analysis of abnormalities in integrity, or connectivity, provided by arcuate fasciculus, fiber bundle involved in language processing, in schizophrenia and VCFS.

Algorithm Component (Golland)

At the core of this project is the stochastic tractography algorithm developed and implemented in collaboration between MIT and BWH. Stochastic Tractography is a Bayesian approach to estimating nerve fiber tracts from DTI images.

We first use the diffusion tensor at each voxel in the volume to construct a local probability distribution for the fiber direction around the principal direction of diffusion. We then sample the tracts between two user-selected ROIs, by simulating a random walk between the regions, based the local transition probabilities inferred from the DTI image.

The resulting collection of fibers and the associated FA values provide useful statistics on the properties of connections between the two regions. To constrain the sampling process to the relevant white matter region, we use atlas-based segmentation to label ventricles and gray matter and to exclude them from the search space. As such, this step relies heavily on the registration and segmentation functionality in Slicer.

Over the last year, we have been working on applying several pre- and postprocessing steps to the algorithm pipeline. These steps include eddy current and geometric distortion correction that have been made available to us by Utah group, as well as DTI filtering (BWH). White matter masks can also now be created based on T2 thresholding within the slicer stochastic tractography module, which makes them more precise, since they do not rely on MRI to DTI co-registration.

At the same time we are working on the datasets where fMRI activations as well as gray matter segmentations need to be registered to DTI data, in order to seed within the predefined gray matter regions. We have made a significant progress in between modality registration, additional improvement is expected when distortion correction become part of the analysis pipeline.

We are also working on improved ways to visualize and quantify stochastic tractography output, not only by parametrizing fiber tracts, but also by creating connection probability distribution maps.

Engineering Component (Davis)

Stochastic Tractography slicer module has been rewritten in python now, and new module released in December 2008, and presented at the AHM in SLC. Its now part of the slicer3. Module documentation have been also created. Current engineering efforts are concentrated on maintaining the module, optimizing it for working with other data formats, and adding new functionality, such as better registration, distortion correction and ways of extracting and measuring FA along the tracts.

Also, because of the fact that the new data is much more computationally demanding (higher spatial resolution, more diffusion directions), and cortical ROIs usually much larger than the previously used WM ROIs, there is general need for performance improvement. This issue is highlighted especially by our stochastic way of tracking connections, where hundreds, instead of just one, (as in deterministic tractography) tracts are being generated from one seed. Thus some of our efforts go towards multithreading, and utilizing parallel processing. Version of our algorithm that uses large computer clusters have been developed and can be downloaded and installed by individual users with minimal knowledge of parallel computing now.

Clinical Component (Kubicki)

Over the last year, we tested the algorithm on newly released 3T NAMIC data, which contains high resolution DTI as well as structural RM data, plus automatic anatomical segmentations. Data is already co-registered, so cortical ROIs can be used as seeding points for stochastic tractography.

Using this dataset, we have completed a clinical study, where we looked at the connections between inferior frontal and superior temporal lobes, sites of the language network. Connections of these two regions, obtained with stochastic tractography, have been measured, and compared between group of 20 chronic schizophrenia patients and 20 controls. We have also looked at gray matter volumes of destination regions, trying to estimate relationship between gray and white matter abnormalities in schizophrenia. Results of this study have been presented at World Psychiatry Congress in Florence, Italy in April of 2009, as well as at Harvard Psychiatry MYSELL conference also in April 2009.

Another clinical study that is under way, is the application of stochastic tractography to define connections involved in emotional processing. For this purpose, we use cortical segmentations of anterior cingulated gyrus, orbital-frontal gyrus and amygdala, and trace as well as quantify connections between there regions in healthy controls as well in schizophrenia patients. Results of this preliminary study have been presented at MYSELL in April 2009, and will be presented at Biological Psychiatry conference later this year.

We are also involved in two collaborative studies. In one, use DTI data acquired in at UCI, and apply stochastic method to segment and measure arcuate fasciculus in subjects with schizophrenia and language impairment, as evinced in ERP data. In another collaboration, we combine resting state fMRI data with DTI in order to measure connectivity between regions forming functional network. Both these projects are under way.

Finally, stochastic tractography have been used qualitatively in one publication that is in press in Human Brain Mapping. Here, we combined fMRI with DTI whole brain data analysis, and found regions that were expressing abnormal functional connectivity in schizophrenia. These regions were then assigned to certain anatomical structures (white mater tracts), based on their location, and relationship to stochastic tractography output.

Additional Information

Additional Information for this project is available here on the NA-MIC wiki.

Roadmap Project: MR-guided Prostate Biopsy Needle Positioning Robot Integration (Fichtinger)

Overview (Fichtinger)

Numerous studies have demonstrated the efficacy of image-guided needle-based therapy and biopsy in the management of prostate cancer. The accuracy of traditional prostate interventions performed using transrectal ultrasound (TRUS) is limited by image fidelity, needle template guides, needle deflection and tissue deformation. Magnetic Resonance Imaging (MRI) is an ideal modality for guiding and monitoring such interventions due to its excellent visualization of the prostate, its sub-structure and surrounding tissues.

We have designed a comprehensive robotic assistant system that allows prostate biopsy and brachytherapy procedures to be performed entirely inside a 3T closed MRI scanner. The current system applies transrectal approach to the prostate: an endorectal coil and steerable needle guide, both tuned to 3T magnets and invariable to any particular scanner, are integrated into the MRI compatible manipulator.

Under the NAMIC initiative, the image computing, visualization, intervention planning, and kinematic planning interface is being accomplished with open source system built on the NAMIC toolkit and its components, such as Slicer3 and ITK. These are complemented by a collection of unsupervised prostate segmentation and registration methods that are of great importance to the clinical performance of the interventional system as a whole.

Algorithm Component (Tannenbaum)

We have worked on both the segmentation and the registration of the prostate from MRI and ultrasound data. We explain each of the steps below.

Prostate Segmentation

We must first extract the prostate. We provided two methods: a shape based method and a semi-automatic method. More details are given below and images and further details may be found here

1. A shape based algorithm. This begins with learning a group of shapes, obtained from manually segmenting a set of prostate 3D images. With the shapes represented as the hyperbolic tangent of the signed distance functions, principle component analysis is employed to learn the shapes. Further, given a new prostate image, we search the learned shape space in order to find one shape best segment the given image. The fitness of one shape to segment the image is evaluated by an energy functional measuring the discrepancy of the statistical characteristics inside and outside the current segmentation boundary. Such method is robust to the noise in the images. Moreover, the whole algorithm pipeline has been integrated into the Slicer3 through the command line module.
2. Semi-automatic method. This method is based on a random walk segmentation algorithm. With user provided initial seed regions inside and out side the object (prostate), the algorithm computes a probability distribution over the image domain by solving a boundary value partial differential equation where the value at seed regions are fixed at 1.0 or 0.0, depending or whether they are object or background seeds. The resulting distribution indicates the probability of each voxel belonging to the object. Simply threshold by 0.5 gives the segmentation of the object. Moreover, if the result is not suitable, the user can edit the seed regions, and the new result is computed based on this previous result. This algorithm has been integrated into the transrectal prostate MRI module of Slier3.

Prostate Registration

We developed a nonlinear (affine) prostate registration method by treating prostate images as point sets. Then the iterative closest point algorithm is improved to register the point sets generated by the two images to be registered. The proposed method shows robustness to long distance transition and partial image structure. Moreover, such representation is much sparser than sampling image on the uniform grid thus the registration is very fast comparing two 3D volumetric image registration.

Furthermore, the registration is viewed as a posterior estimation problem, in which the distributions of the affine and translation parameters are to be estimated. This can naturally be estimated using a particle filter framework. Through this, the method can handle the otherwise difficult cases where the two prostates are one supine and one prone.

More details are given here...

Engineering Component (Hayes)

<Note Progress in the last year>

Clinical Component (Fichtinger)

<Note Progress in the last year>

Additional Information

Additional Information for this project is available here on the NA-MIC wiki.

Roadmap Project: Brain Lesion Analysis in Neuropsychiatric Systemic Lupus Erythematosus (Bockholt)

Overview (Bockholt)

The primary goal of the MIND DPB is to examine changes in white matter lesions in adults with Neuropsychiatric Systemic Lupus Erythematosus (SLE). We want to be able to characterize lesion location, size, and intensity, and would also like to examine longitudinal changes of lesions in an SLE cohort. To accomplish this goal, we will create an end-to-end application entirely within NA-MIC Kit allowing individual analysis of white matter lesions. Such a workflow will then be applied to a clinical sample in the process of being collected.

Algorithm Component (Whitaker)

The basic steps necessary for the white matter lesion analysis application entail first registration of T1, T2, and FLAIR images, second tissue classification into gray, white, csf, or lesion, thirdly clustering lesion for anatomical localization, and finally a summarization of lesion size and image intensity parameters within each unique lesion.

<Note Progress in the last year>

Engineering Component (Pieper)

<Note Progress in the last year>

Clinical Component (Bockholt)

<Note Progress in the last year>

Additional Information

Additional Information for this project is available here on the NA-MIC wiki.

Roadmap Project: Cortical Thickness for Autism(Hazlett)

Overview (Hazlett)

A primary goal of the UNC DPB is to examine changes in cortical thicknes in children with autism compared to typical controls. We want to examine group differences in both local and regional cortical thickness, and would also like to examine longitudinal changes in the cortex from ages 2-4 years. To accomplish this goal, this project will create an end-to-end application within Slicer3 allowing individual and group analysis of regional and local cortical thickness. Such a workflow will then be applied to our study data (already collected).

Algorithm Component (Styner)

The basic steps necessary for the cortical thickness application entail first tissue segmentation in order to separate white and gray matter regions, second cortical thickness measurement, thirdly cortical correspondence to compare measurements across subjects and finally a statistical analysis to locally compute group differences.

<Note Progress in the last year>

Engineering Component (Miller, Vachet)

<Note Progress in the last year>

Clinical Component (Hazlett)

<Note Progress in the last year>

Additional Information

Additional Information for this project is available here on the NA-MIC wiki.

Four Infrastructure Topics

Diffusion Image Analysis (Gerig)

<Note Progress in the last year>

Key Investigators

<Need to update the list below>

• BWH: Marek Kubicki, Martha Shenton, Sylvain Bouix, Julien von Siebenthal, Thomas Whitford, Jennifer Fitzsimmons, Doug Terry, Jorge Alverado, Eric Melonakos, Carl-Fredrik Westin.
• MIT: Lauren O'Donnell, Polina Golland
• UCI: James Fallon, Judi Ford
• Utah I: Tom Fletcher, Ross Whitaker, Ran Tao, Yongsheng Pan
• Utah II: Casey Goodlett, Sylvain Gouttard, Guido Gerig
• GA Tech: John Melonakos, Vandana Mohan, Shawn Lankton, Allen Tannenbaum
• GE: Xiaodong Tao, Jim Miller, Mahnaz Mandah
• Isomics: Steve Pieper
• Kitware: Luis Ibanez

Additional Information

Additional Information for this topic is available here on the NA-MIC wiki.

Structural Analysis(Tannenbaum)

Progress

Under Structural Analysis, the main topics of research for NAMIC are structural segmentation, registration techniques and shape analysis. These topics are correlated and hence research in one often finds application in another. For example, shape analysis can yield useful priors for segmentation, or segmentation and registration can provide structural correspondences for use in shape analysis and so on.

An overview of selected progress highlights under these broad topics follows:

Segmentation

• Geodesic Tractography Segmentation: We proposed an image segmentation technique based on augmenting the conformal (or geodesic) active contour framework with directional information. This has been applied successfully to the segmentation of neural fiber bundles such as the Cingulum Bundle. This framework has now been integrated into Slicer and is being tested on a population of brain data sets.

• Tubular Surface Segmentation: We have proposed a new model for tubular surfaces that transforms the problem of detecting a surface in 3D space, to detecting a curve in 4D space. Besides allowing us to impose a "soft" tubular shape prior, this also leads to computational efficiency over conventional surface segmentation approaches. We have also developed the moving end points implementation of this framework wherein the required input is only a few points in the interior of the structure of interest. This yields the additional advantage that the framework simulatenously returns both the 3D segmentation and the 3D skeleton of the structure eliminating the need for apriori knowledge of end points, and an expensive skeletonization step. The framework is applicable to different tubular anatomical structures in the body. We have so far applied it successfully to the Cingulum Bundle, and blood vessels.

• Local-global Segmentation: We have proposed a novel segmentation approach that combines the advantages of local and global approaches to segmentation, by using statistics over regions that are local to each point on the evolving countour. This makes it well suited to applications with contrast differences within the structure of interest such as in blood vessel segmentation, as well as applications like the neural fiber bundles where the diffusion profiles of voxels within the structure are locally similar but vary along the length of the fiber bundle itself.

• Shape-based segmentation: Standard image based segmentation approaches perform poorly when there is little or no contrast along boundaries of different regions. In such cases segmentation is mostly performed manually using prior knowledge of the shape and relative location of the underlying structures combined with partially discernible boundaries. We have presented an automated approach guided by covariant shape deformations of neighboring structures, which is an additional source of prior knowledge. Captured by a shape atlas, these deformations are transformed into a statistical model using the logistic function. The mapping between atlas and image space, structure boundaries, anatomical labels, and image inhomogeneities are estimated simultaneously within an Expectation-Maximization formulation of the Maximum A posteriori Probability (MAP) estimation problem. These results are then fed into an Active Mean Field approach, which views the results as priors to a Mean Field approximation with a curve length prior. We have applied the algorithm successfully to real MRI images, and we have also implemented it into 3D Slicer.

• Re-Orientation Approach for Segmentation of DW-MRI: This work proposes a methodology to segment tubular fiber bundles from diffusion weighted magnetic resonance images (DW-MRI). Segmentation is simplified by locally reorienting diffusion information based on large-scale fiber bundle geometry. Segmentation is achieved through simple global statistical modeling of diffusion orientation which allows for a convex optimization formulation of the segmentation problem, combining orientation statistics and spatial regularization. The approach compares very favorably with segmentation by full-brain streamline tractography.

Registration

• Optimal Mass Transport based Registration: We have provided a computationaly efficient non-rigid/elastic image registration algorithm based on the Optimal Mass Transport theory. We use the Monge-Kantorovich formulation of the Optimal Mass Transport problem and implement the solution proposed by Haker et al. using multi-resolution and multigrid techniques to speed up the convergence. We also leverage the computation power of general purpose graphics processing units available on standard desktop computing machines to exploit the inherent parallelism in our algorithm. We extend the work by Haker et al. who compute the optimal warp from a first order partial differential equation, an improvement over earlier proposed higher order methods and those based on linear programming, and further implement the algorithm using a coarse-to-fine strategy resulting in phenomenol improvement in convergence. We have applied it successfully to the registration of 3D brain MRI datasets (preoperative and intra-operative), and are currently extending it to the non-rigid registration of baseline DWI to brain MRI data.

• Atlas Regularization for Image Segmentation: Atlas-based approaches have demonstrated the ability to automatically identify detailed brain structures from 3-D magnetic resonance (MR) brain images. Unfortunately, the accuracy of this type of method often degrades when processing data acquired on a different scanner platform or pulse sequence than the data used for the atlas training. In this paper, we improve the performance of an atlas-based whole brain segmentation method by introducing an intensity renormalization procedure that automatically adjusts the prior atlas intensity model to new input data. Validation using manually labeled test datasets has shown that the new procedure improves the segmentation accuracy (as measured by the Dice coefficient) by 10% or more for several structures including hippocampus, amygdala, caudate, and pallidum. The results verify that this new procedure reduces the sensitivity of the whole brain segmentation method to changes in scanner platforms and improves its accuracy and robustness, which can thus facilitate multicenter or multisite neuroanatomical imaging studies.

• Point-set Rigid Registration: We have proposed a particle filtering scheme for the registration of 2D and 3D point set undergoing a rigid body transformation. Moreover, we incorporate stochastic dynamics to model the uncertainity of the registration process. We treat motion as a local variation in the pose parameters obatined from running a few iterations of the standard Iterative Closest Point (ICP) algorithm. Employing this idea, we introduce stochastic motion dynamics to widen the narrow band of convergence as well as provide a dynamical model of uncertainity. In contrast with other techniques, our approach requires no annealing schedule, which results in a reduction in computational complexity as well as maintains the temoral coherency of the state (no loss of information). Also, unlike most alternative approaches for point set registration, we make no geometric assumptions on the two data sets.We applied the algorithm to different alignments of point clouds and it successfully found the correct optimal transformation that aligns two given point clouds despite the differing geometry around the local neighborhood of a point within their respective sets.

• Regularization for Optimal Mass Transport: To extend the flexibility of the existing OMT algorithm, we added a regularization term to the functional being minimized. This term controls the tradeoff between how well two images match after registration versus how warped the transformation map can become. A weighted sum of squared differences is used to penalize having to move mass over long distances; this addition also helps to keep the transformation physically accurate by reducing the likelihood that the transformation grid will fold over itself and keeping the grid smooth.

• Registration of DW-MRI to structural MRI: Optimal Mass Transport was applied to the problem of correcting EPI distortion in DW-MRI. A mask for white matter in DW-MRI was registered to the white matter mask extracted from the structural MRI for the same patient. Prior to registration, it is important to normalize intensities in the two masks; this was done by dividing the images into regions and uniformly normalizing over each region to assure the sum of the intensities is equal. Then, once a transformation between the white matter masks was calculated, this transformation was applied to the original DW-MRI image.

Shape Analysis

• Shape Analysis Framework using SPHARM-PDM: We have provided an analysis framework of objects with spherical topology, described by sampled spherical harmonics SPHARM-PDM. The input is a set of binary segmentations of a single brain structure, such as the hippocampus or caudate. These segmentations are first processed to fill any interior holes. The processed binary segmentations are converted to surface meshes, and a spherical parametrization is computed for the surface meshes using a area-preserving, distortion minimizing spherical mapping. The SPHARM description is computed from the mesh and its spherical parametrization. Using the first order ellipsoid from the spherical harmonic coefficients, the spherical parametrizations are aligned to establish correspondence across all surfaces. The SPHARM description is then sampled into a triangulated surfaces (SPHARM-PDM) via icosahedron subdivision of the spherical parametrization. These SPHARM-PDM surfaces are all spatially aligned using rigid Procrustes alignment. Group differences between groups of surfaces are computed using the standard robust Hotelling T 2 two sample metric. Statistical p-values, both raw and corrected for multiple comparisons, result in significance maps. We provide additional visualization of the group tests via mean difference magnitude and vector maps, as well as maps of the group covariance information. We have a stable implementation, and current development focuses on integrating the current command line tools into Slicer (v3) via the Slicer execution model.

• Population studies using Tubular Surface Model: We have proposed a tubular shape model for the Cingulum Bundle which models a tubular surface as a center-line coupled with a radius function at every point along the center-line. This model shows potential for population studies on the Cingulum Bundle which is believed to be involved in Schizophrenia, since it provides a natural way of sampling the structure to build a feature representation of it. We are currently segmenting the Cingulum Bundle from a population of brain data sets, towards performing this population analysis using the Pott's Model.

• Automatic Outlining of Sulci on a Brain Surface: We present a method to automatically extract certain key features on a surface. We apply this technique to outline sulci on the cortical surface of a brain, where the data is taken to be a 3D triangulated mesh formed from the segmentation of MR image slices. The problem is posed as energy minimization using penalizing the arc-length of segmenting curve using conformal factor involving the mean curvature of the underlying surface. The computation is made practical for dense meshes via the use of a sparse-field method to track the level set interfaces and regularized least-squares estimation of geometric quantities.

Key Investigators

Needs to be updated:

• MIT: Polina Golland, Kilian Pohl, Sandy Wells, Eric Grimson, Mert R. Sabuncu
• UNC: Martin Styner, Ipek Oguz, Xavier Barbero
• Utah: Ross Whitaker, Guido Gerig, Suyash Awate, Tolga Tasdizen, Tom Fletcher, Joshua Cates, Miriah Meyer
• GaTech: Allen Tannenbaum, John Melonakos, Vandana Mohan, Tauseef ur Rehman, Shawn Lankton, Samuel Dambreville, Yi Gao, Romeil Sandhu, Xavier Le Faucheur, James Malcolm, Ivan Kolosev
• Isomics: Steve Pieper
• GE: Bill Lorensen, Jim Miller
• Kitware: Luis Ibanez, Karthik Krishnan
• UCLA: Arthur Toga, Michael J. Pan, Jagadeeswaran Rajendiran
• BWH: Sylvain Bouix, Motoaki Nakamura, Min-Seong Koo, Martha Shenton, Marc Niethammer, Jim Levitt, Yogesh Rathi, Marek Kubicki, Steven Haker

Additional Information

Additional Information for this topic is available here on the NA-MIC wiki.

fMRI Analysis (Golland)

Progress

One of the major goals in analysis of fMRI data is the detection of functionally homogeneous networks in the brain.

<note progress here>

Key Investigators

Need to update this list:

1. MIT: Polina Golland, Danial Lashkari, Bryce Kim
2. Harvard/BWH: Sylvain Bouix, Martha Shenton, Marek Kubicki

Additional Information

Additional Information for this topic is available here on the NA-MIC wiki.

NA-MIC Kit Theme (Schroeder)

Progress

The NAMIC-Kit consists of a framework of advanced computational components, as well as the support infrastructure for testing, documenting, and deploying leading edge medical imaging algorithms and software tools. The framework has been carefully constructed to provide low-level access to libraries and modules for advanced users, plus high-level application access that non-computer professionals can use to address a variety of problems in biomedical computing. In this fifth year of the NA-MIC projects <summary of progress>

Software Releases

The NAMIC-Kit can be represented as a pyramid of capabilities, with the base consisting of toolkits and libraries, and the apex standing in for the Slicer3 user application. In between, Slicer modules are stand-alone executables that can be integrated directly into the Slicer3 application, including GUI integration, while work-flows are groups of modules that are integrated together to manifest sophisticated segmentation, registration and biomedical computing algorithms. In a coordinated NAMIC effort, major releases of these many components were realized over the past year. This includes, but is not limited to:

Slicer3 and the Software Framework

One of the major achievements of the past year has been...

Software Process

One of the challenges facing developers has been the requirement to implement, test and deploy software systems across multiple computing platforms. NAMIC continues to push the state of the art with further development of the CMake, CTest, and CPack tools for cross-platform development, testing, and packaging, respectively...

Key Investigators

THis list needs to be updated:

• Kitware - Will Schroeder (Core 2 PI), Sebastien Barre, Luis Ibanez, Bill Hoffman
• GE - Jim Miller, Xiaodong Tao
• Isomics - Steve Pieper

Additional Information

Additional Information for this topic is available here on the NA-MIC wiki.

Highlights(Schroeder)

Outreach and Technology Transfer

Cores 4-5-6 continue to support, train and dissemniate to the NAMIC community, and the broader biomedical computing community.

• The Slicer community held several workshops and tutorials. In xxx a satellite event was held for the international Organization for Human Brain Mapping at the annual meeting in xxx. The xx workshop on xx hosted xx participants representing xx countries from around the world, xx states within the US and xxdifferent laboratories including xx NIH institutes. In addition, <note how many slicer tutorials were held and where etc>
• Project Week continues to be a successful NAMIC venue. These semi-annual events are held in Boston in June, and January in Salt Lake City. These events are well attended with approximately 100 participants, of which about a third are outside collaborators. At the last Project Week in Salt Lake City, approximately xx projects were realized.
• NAMIC continues to participate in conferences and other technical venues. For example, NAMIC hosted xxx

Impact and Value to Biocomputing (Miller)

NA-MIC impacts Biocomputing through a variety of mechanisms. First, NA-MIC produces scientific results, methodologies, workflows, algorithms, imaging platforms, and software engineering tools and paradigms in an open enviroment that contributes directly to the body of knowledge available to the field. Second, NA-MIC science and technology enables the entire medical imaging community to build on NA-MIC results, methods, and techniques, to concentrate on the new science instead of developing supporting infrastructure, to leverage NA-MIC scientists and engineers to adapt NA-MIC technology to new problem domains, and to leverage NA-MIC infrastructure to distribute their own technology to a larger community.

Timeline (Ross)

<The table needs to be updated>

This section of the report gives the milestones for years 1 through 5 that are associated with the timelines in the original proposal. We have organized the milestones by core. For each milestone we have indicated the proposed year of completion and a very brief description of the current status. In some cases the milestones include ongoing work, and we have try to indicate that in the status. We have also included tables that list any significant changes to the proposed timelines. On the wiki page, we have links to the notes from the various PIs that give more details on their progress and the status of the milestones.

These tables demonstrate that the project is, on the whole, proceeding according to the originally planned schedule.

Core 1: Algorithms

Timelines and Milestones

Group	Aim	Milestone	Proposed time of completion	Status
MIT	1	Shape-based segmentation
MIT	1.1	Methods to learn shape representations	Year 2	Completed
MIT	1.2	Shape in atlas-driven segmentation	Year 4	Completed
MIT	1.3	Validate and refine approach	Year 5	In Progress
MIT	2	Shape analysis
MIT	2.1	Methods to compute statistics of shapes	Year 4	Completed
MIT	2.3	Validation of shape methods on application data	Year 5	Completed, refinements ongoing
MIT	3	Analysis of DTI data
MIT	3.1	Fiber geometry	Year 3	Completed
MIT	3.2	Fiber statistics	Year 5	Completed, new developments ongoing
MIT	3.3	Validation on real data	Year 5	Completed, refinements ongoing
Utah	1	Processing of DTI data
Utah	1.1	Filtering of DTI	Year 2	Completed
Utah	1.2	Quantitative analysis of DTI	Year 3	Completed, refinements ongoing
Utah	1.3	Segmentation of cortex/WM	Year 3	Completed partially, modified below
Utah	1.4	Segmentation analysis of white matter tracts	Year 3	Completed, applications ongoing
Utah	1.5	Joint analysis of DTI and functional data	Year 5	Initiated
Utah	2	Nonparametric Shape Analysis	Year 5	Completed
Utah	2.1	Framework in place	Year 3	Complete
Utah	2.2	Demonstration on shape of neuranatomy (from Core 3)	Year 4	Complete
Utah	2.3	Development for multiobject complexes	Year 4	Complete
Utah	2.4	Demonstration of NP shape representations on clinical hypotheses from Core 3	Year 5	Complete, publications in progress
Utah	2.6	Integration into NAMIC-kit	Year 5	Incomplete (initiated)
Utah	2.7	Shape regression	Year 5	Incomplete
UNC	1	Statistical shape analysis
UNC	1.1	Comparative anal. of shape anal. schemes	Year 2	Completed
UNC	1.3	Statistical shape analysis incl. patient variable	Year 5	Complete, refinements ongoing
UNC	2	Structural analysis of DW-MRI
UNC	2.1	DTI tractography tools	Year 4	Completed
UNC	2.2	Geometric characterization of fiber tracts	Year 5	Completed
UNC	2.3	Quant. anal. of diffusion along fiber tracts	Year 5	Completed.
GaTech	1.1	ITK Implementation of PDEs	Year 2	Completed
GaTech	1.1	Applications to Core 3 data	Year 4	Completed
GaTech	1.2	New statistic models	Year 4	Completed
GaTech	1.2	Shape anaylsis	Year 4	Completed, refinements ongoing
GaTech	2.0	Integration in to Slicer	Year 4-5	Preliminary results and ongoing
MGH	1	Registration
MGH	1.1	Collect DTI/QBALL data	Year 2	Completed
MGH	1.2	Develop registration method	Year 2	Completed
MGH	1.3	Test/optimize registration method	Year 3	In Progress
MGH	1.4	Apply registration on core 3 data	Year 5	In Queue
MGH	2	Group DTI Statistics
MGH	2.1	Develop group statistic method	Year 2	Partially Complete
MGH	2.2	Apply on core 3 data	Year 5	In Queue
MGH	3	Diffusion Segmentation
MGH	3.1	Collect DTI/QBALL data	Year 2	Completed
MGH	3.2	Develop/optimize segmentation algorithm	Year 3	Modified
MGH	3.3	Integrate w/ tractography	Year 4	Modified
MGH	3.4	Apply on core 3 data	Year 5	Modified
MGH	4	Group Morphometry Statistics
MGH	4.1	Develop/optimize statistics algorithms	Year 3	Modified
MGH	4.2	Develop GUI for Linux	Year 3	Modified
MGH	4.3	Slicer integration	Year 3	Modified
MGH	4.4	Compile application on Windows	Year 4	Modified
MGH	5	XNAT Desktop	Years 4-5
MGH	5.1	Establish requirements for desktop version of XNAT	Years 4-5	Complete
MGH	5.2	Develop implementation plan for prototype	Years 4-5	Complete
MGH	5.3	Implement prototype version	Years 4-5	Incomplete (in progress)
MGH	5.4	Implement alpha version	Year 5	Incomplete
MGH	6	XNAT Central	Years 4-5
MGH	6.1	Deploy XNAT Central, a public access XNAT host	Years 4-5	Complete
MGH	6.2	Coordinate with NAMIC sites to upload project data	Years 4-5	Incomplete (ongoing)
MGH	6.3	Continue developing XNAT Central based on feedback from NAMIC sites	Years 4-5	Incomplete (ongoing)
MGH	7	NAMIC Kit integration	Years 4-5
MGH	7.1	Implement web services to exchange data with Slicer, Batchmake, and other client applications	Years 4-5	Incomplete (ongoing)
MGH	7.2	Add XNAT Desktop to standard NAMIC kit distribution	Year 5	Incomplete

Timeline Modifications

Group	Aim	Milestone	Modification
MIT	2.2	Methods to compare shape statistics	Removed, the effort refocused on registration necessary for population studies
MIT	2.4	Software infrastructure to integrate shape analysis tools into the pipeline for population studies.	New, morphed into collaboration with XNAT to provide more general population analysis tools. Partially completed.
MIT	4	fMRI analysis including local and atlas-based priors for quantifying activation.	New, partially completed. Refinements in progress. Clinical study with Core 1 is in progress.
Utah	2.2 (removed)	Feature-based brain image registration.	Shift emphasis to shape-based analysis/registration
Utah	2.1 (removed)	Cortical filtering and feature detection	Effort is subsumed by other Core 1 partners (e.g. see MGH/Freesurfer)
Utah	1.3 (removed)	Segmentation of cortex/WM	Effort is subsumed by other Core 1-2 partners (e.g. see EM-Segmenter)
Utah	3.0 (removed)	Fast implmentations of PDEs	Real-time filtering is demphasized in favor of shape/DTI analysis
Utah	1.5 (added)	Joint analysis of DTI and functional data	Opportunities/needs within various collaborations
Utah	2.1-2.3 (added, in place of cortical analysis)	Shape analysis	Nonparametric shape analysis added to address needs of core 3.
Utah	2.7	Shape regression	Extension/completion of framework. Opportunities/needs within various collaborations.
UNC	1.2	Develop medially-based shape representation	Remove
UNC	1.4	Develop generic cortical correspondence framework (Years 3-5)	New
UNC	2.4	DTI Atlas Building (Years 2--4)	New
GaTech	2.1	FA analysis	New
MGH	4.1 - 4.4	Group Morphometry Statistics	Added and then removed, based on personnel changes
MGH	5-7	XNAT	Added to support remote image database capabilities

Core 2: Engineering

Core 2 Timelines and Milestones

Group	Aim	Milestone	Proposed time of completion	Status
GE	1	Define software architecture
GE	1	Object design	Yr 1	Completed
GE	1	Identify patterns	Yr 3	Patterns for processing scalar and vector images, models, fiducials complete. Patterns for diffusion weighted completed, fMRI ongoing.
GE	1	Create frameworks	Yr 3	Frameworks for processing scalar and vector images, models, fiducials complete. Frameworks for diffusion weighted completed, fMRI ongoing.
GE	2	Software engineering process
GE	2	Extreme programming	Yr 1-5	On schedule, ongoing
GE	2	Process automatiion	Yr 3	On schedule, ongoing
GE	2	Refactoring	Yr 3	Complete
GE	3	Automated quality system
GE	3	DART deployment	Yr 2	Complete
GE	3	Persistent testing system	Yr 5	Incomplete
GE	3	Automatic defect detection	Yr 5	Incomplete
Kitware	1	Cross-platform development
Kitware	1	Deploy environment (CMake, CTest)	Yr 1	Complete
Kitware	1	DART Integration and testing	Yr 1	Complete
Kitware	1	Documentation tools	Yr 2	Complete
Kitware	2	Integration tools
Kitware	2	File Formats/IO facilities	Yr 2	Complete
Kitware	2	CableSWIG deployment	Yr 3	Complete (integration ongoing)
Kitware	2	Establish XML schema	Yr 4	Complete, refinements ongoing
Kitware	3	Technology delivery
Kitware	3	Deploy applications	Yr 1	Complete (ongoing)
Kitware	3	Establish plug-in repository	Yr 2	Incomplete
Kitware	3	Cpack	Yr 4-5	Incomplete
Isomics	1	NAMIC builds of slicer	Years 2--5	Complete
Isomics	1	Schizophrenia and DBP intefaces	Year 3---5	Completed (refinements ongoing)
Isomics	2	ITK Integration tools	Year 1---3	Completed
Isomics	2	Experiment Control Interfaces	Year 2---5	Migration from LONI to BatchMake Underway
Isomics	2	fMRI/DTI algorithm support	Year 2---5	Completed DTI, fMRI Ongoing
Isomics	2	New DBP algorithm support	Year 2---5	Ongoing
Isomics	3	Compatible build process	Year 1---3	Completed
Isomics	3	Dart Integration	Year 1---2	Completed (upgrades ongoing)
Isomics	3	Test scripts for new code	Year 2---5	Ongoing
UCSD	1	Grid computing---base	Year 1	Completed
UCSD	1	Grid enabled algorithms	Year 3	First version (GWiz alpha) available - initial integration with Slicer3 and execution model.
UCSD	1	Testing infrastructure	Year 4	Initiated
UCSD	2	Data grid --- compatibility	Year 2	Completed
UCSD	2	Data grid --- slicer access	Year 2	Completed for version 2.6. In progress for Slicer3
UCSD	3	Data mediation --- deploy	Year 1	Incomplete (modfication below)
UCLA	1	Debabeler functionality	Year 1	Continued Progress
UCLA	2	SLIPIE Interpretation (Layer 1)	Year 1--Year2	In Progress
UCLA	3	SLIPIE Interpretation (Layer 2)	Year 1--Year2	On Schedule
UCLA	3	Developing ITK Modules	Year2	In Progress
UCLA	4	Integrating SRB (GSI-enabled)	Year2	Completed
UCLA	5	Integrating IDA	Year2	Completed
UCLA	5	Integrating External Visualization Applications	Year2	Completed

Core 2 Timeline Modifications

Group	Aim	Milestone	Modification
Isomics	3	Data mediation	Delayed pending integration of databases into NAMIC infractructure

Core 3: Driving Biological Problems

The Core 3 projects submitted R01 style proposals, as specified in the RFA, and did not submit timelines.

Core 4: Service

Core 4 Timelines and Milestones

Group	Aim	Milestone	Proposed time of completion	Status
Kitware	1	Implement Development Farms
Kitware	1	Deploy platforms	Yrs 1	Complete
Kitware	1	Communications	Yrs 1	Complete, ongoing
Kitware	2	Establish software process
Kitware	2	Secure developer database	Yr 1	Complete, ongoing
Kitware	2	Collect guidelines	Yr 1	Complete
Kitware	2	Manage software submission process	Yr 1	Complete
Kitware	2	Configure process tools	Yr 1	Complete
Kitware	2	Survey community	Yr 1	Complete
Kitware	3	Deploy NAMIC Tools
Kitware	3	Toolkits	Yr 1	Complete
Kitware	3	Integration tools	Yr 1	Complete
Kitware	3	Applications	Yr 1	Complete
Kitware	3	Integrate new computing resources	Yr 1	Complete
Kitware	4	Provide support
Kitware	4	Esablish support infrastructure	Yrs 1--5	On schedule, ongoing
Kitware	4	NAMIC support	Yr 1	Complete
Kitware	5	Manage NAMIC Software Releases	Yrs 1--5	On schedule, ongoing

Core 4 Timeline Modifications

Group	Aim	Milestone	Modification
Kitware	2-5	Various	Refined/modified the sub aims

Core 5: Training

Core 5 Timelines and Milestones

Group	Aim	Milestone	Proposed time of completion	Status
Harvard	1	Formal Training Guidllines
Harvard	1	Functional neuroanatomy	Yr 1	Complete
Harvard	1	Clinical correlations	Yr 1	Complete
Harvard	2	Mentoring
Harvard	2	Programming workshops	Yrs 1-5	On schedule, ongoing
Harvard	2	One-on-one mentoring, Cores 1, 2, 3	Yrs 1-5	On schedule, ongoing
Harvard	3	Collaborative work environment
Harvard	3	Wiki	Yrs 1	Complete
Harvard	3	Mailing lists	Yrs 1	Complete
Harvard	3	Regular telephone conferences	Yrs 1-5	On schedule, ongoing
Harvard	4	Educational component for tools
Harvard	4	Slicer training modules	Yr 2-5	Slicer 2.x tutorials complete, Two Slicer 3 tutorials complete, translation of 2.x tutorials to 3 is ongoing and on schedule
Harvard	5	Demonstrations and hands-on training
Harvard	5	Various workshops and conferences	Yrs 1--5	On schedule, ongoing

Core 5 Timeline Modifications

None.

Core 6: Dissemination

Core 6 Timelines and Milestones

Group	Aim	Milestone	Proposed time of completion	Status
Isomics	1	Create a collaboration metholdology for NA-MIC
Isomics	1.1	develop a selection process	Yr 1	Complete
Isomics	1.2	guidelines to govern the collaborations	Yr 1-2	Complete
Isomics	1.3	Provide on-site training	Yr 1-5	Complete for current tools (ongoing for tool refinement)
Isomics	1.4	develop a web site infrastructure	Yr 1	Complete
Isomics	2	Facilitate communication between NA-MIC developers and wider research community
Isomics	2.1	develop materials describing NAMIC technology	Yr 1-5	On Schedule
Isomics	2.2	participate in scientific meetings	Yr 2-5	On Schedule
Isomics	2.3	Document interactions with external researchers	Yr 2-5	On Schedule
Isomics	2.4	Coordinate publication strategies	Yr 3-5	On Schedule
Isomics	3	Develop a publicly accessible internet resource of data, software, documentation, and publication of new discoveries
Isomics	3.1	On-line repository of NAMIC related publications and presentations	Yr 1-5	On Schedule
Isomics	3.2	On-line repository of NAMIC tutorial and training material	Yr 1-5	On Schedule
Isomics	3.3	Index and a searchable database	Yr 1-2	Done
Isomics	3.4	Automated feedback systems that track software downloads	Yr 3	Done

Core 6 Timeline Modifications

None.

Appendix A Publications (Mastrogiacomo)

A list should be mined from the publications database and attached here in MS word format.