ANALYSIS OF MULTI-VOLUME SEQUENCES OF MEDICAL IMAGING DATA
Open source software for medical image analysis:
3D Slicer (or simply "Slicer") is a software package for medical image registration, segmentation, analysis and
visualization (See www.slicer.org and www.na-mic.org's Slicer Wiki for more information). Slicer is free and open-source, with cross-platform support, high-performance C++ image
analysis and an interactive, multi-modality visualization environment; it is an extensible platform for research
and development with a consistent graphical interface and coding convention; it has a GUI-based front-end but
can also be used for grid computing batch jobs. Slicer is used both as a vehicle for delivering multi-modality medical image
analysis and visualization tools to clinical investigators worldwide, and as a programming environment for the development
and testing of new medical image computing techniques. Its development has historically been a collaboration
between the MIT Artificial Intelligence Lab and the Surgical Planning Lab at Brigham and Women's Hospital, Harvard
Medical School.
Multi-volume image analysis and visualization: To support the research programs at the HCNR and
fBIRN , colleagues Steven Pieper, Sandy Wells,
Haiying Liu, Cindy Wible, Carsten Richter and I are developing a set of software modules within the Slicer environment
to: load and inspect large multi-volume sequences of medical data; pre-process them for analysis; perform statistical
analysis using conventional and emerging techniques; and visualize the results. Our main focus in on the analysis and
visualization of functional Magnetic Resonance Imaging (fMRI) data, though many of the tools are applicable to other
applications that also require the analysis of multi-volume sequences of data, such as cardiac perfusion imaging, image-guided
surgery, and the longitudinal study of disease and disease processes.
functional MRI: We are developing the fMRIEngine, a Slicer software module for detecting and visualizing brain activations in fMRI data. This module allows multi-volume fMRI acquisitions to be loaded or imported after pre-processing from the Interval Browser. A description of the experimental paradigm used during the acquisition can be specified through the fMRIEngine's GUI panel. Additional signal modeling may be applied to the basic waveforms that encode the experimental conditions in the paradigm, such as convolution with a hemodynamic response function or the addition of temporal derivatives to approximate response latency. The fMRIEngine is designed to support an extensible suite of brain activation detection techniques; currently conventional linear modeling for first level analyses has been implemented. Once the model and contrasts of parameter estimates have been specified, the "surfable" design matrix can be inspected as shown below.
Visualizing and inspecting results: This Resulting volumes of brain activation statistics can be thresholded and visualized using a color map; the map represents the statistical significance of each voxel's activation under conditions of interest in the paradigm. Currently, we are developing tools within the fMRIEngine to perform analysis within regions of interest, defined by underlying anatomy by "blobs" of activated voxels. Voxel timecourses can also be plotted by interactively clicking on a voxel within Slicer's "Slice" shown beneath the 3D view.
Interval Browser: The Interval Browser (or Ibrowser) module provides a GUI for organizing, inspecting and animating sequences of multi-volume data along horizontal tracks, or "intervals". Each interval contains icons that act as shortcuts for configuring visibility and performing common tasks like copying and deleting, and specifying interval order. The GUI also provides icons for animation, viewing and manual indexing of volumes in each interval. The visual interface, shown below, provides an important tool for managing large collections of volumetric data within 3D Slicer.
In the standard tabbed GUI panel Slicer provides to each of its software modules, the Ibrowser also includes features for loading, processing and inspecting multi-volume sequences. Its evolving suite of processing tools is designed to be extensible, and to accommodate application-specific processing workflows. Like other Slicer modules, the Ibrowser is implemented using VTK/C++ with a Tcl/Tk wrapper.
MRI cardiac perfusion: Techniques for MRI cardiac perfusion analysis also operate on multi-volume data sequences and consider each voxel timecourse in the analysis. Similarly to results from fMRI brain activation detection, results of the perfusion analysis are typically viewed as color-coded parametric maps overlayed onto a high-resolution anatomical scan. And as in fMRI pre-processing, each acquired sequence must be motion corrected to ensure that all voxels within a timecourse represent the same spatial anatomical location. The nature of cardiac perfusion data presents additional challenges for automated registration techniques due to the large deformations from physiological motion, and the large localized changes in image intensity due to the contrast bolus. Currently I'm collaborating with colleagues Steven Pieper and Raymond Kwong M.D. at Brigham and Women's Hospital on extending Slicer's multi-volume infrastructure to support the analysis of MRI cardiac pefusion data.
