Diffusion tensor imaging (DTI) allows a quantitative study of the
underlying geometry and architecture of white matter. We have
developed a suite of open source tools for quantitative analysis of
DTI suitable for use in clinical research studies. We have leveraged
open source tools for file IO, image processing, visualization, and
user interface. The tools are designed to provide several
interconnected modules to perform quantitative analysis of DTI data
based on fiber tracts generated from tractography.

A Quantitative DTI Fiber Tract Analysis Suite

This paper investigates an Open Framework for Medical Simulation (named SOFA for Simulation Open Framework Architecture), and proposes ideas for structuring and supporting its expansion. Lately, through a joint effort between the Sim Group at CIMIT and the Alcove group at INRIA, we have investigated the foundations for a more extensive, flexible, and comprehensive framework. In addition, five internationally renowned research groups have contributed to this initiative by converting their algorithms into compatible modules. Although still in its very early developmental phase, the current project illustrates some of the key concepts we believe will enable collaboration and interoperability in Medical Simulation. The main objective of SOFA research is to foster collaboration among research groups. It is our hope that SOFA will simplify the development cycles, reduce production costs, and provide a means to share components through a common interface.

Collaborative Development of an open framework for medical simulation

An Insight Toolkit (ITK) implementation of our knowledgebased
segmentation algorithm applied to brain MRI scans is presented
in this paper. Our algorithm is a refinement of the work of Teo, Saprio,
and Wandall. The basic idea is to incorporate prior knowledge into the
segmentation through Bayesrule. Image noise is removed via an affine
invariant anisotropic smoothing of the posteriors as in Haker et. al.
We present the results of this code on two different projects. First, we
show the effect of applying this code to skull-removed brain MRI scans.
Second, we show the effect of applying this code to the extraction of the
DLPFC from a user-defined subregion of brain MRI data.We present our
results on brain MRI scans, comparing the results of the knowledge-based
segmentation to manual segmentations on datasets of schizophrenic patients.

Knowledge-Based Segmentation of Brain MRI Scans Using the Insight Toolkit

We have successfully created a software environment in which ultrasound data can be manipulated by, ITK (the Insight Tool-Kit), in real-time. We were able to access each frame generated within the resident computer of a TerasonTM Ultrasound Machine, convert it into the ITK image format, and demonstrate the concurrent operation of ITK on the same computer by writing the images to an external hard drive. At a rate of 10 frames per second, 512 by 512 pixel grayscale frames were written by ITK methods to the external hard drive through USB 2.0 while the ultrasound scan was occurring without thrashing or delay in system performance. This simple exercise demonstrates the potential of ITK in processing ultrasound images in real-time in addition to the more traditional off-line processing.

Real-Time Ultrasound Image Analysis for the Insight Toolkit

We have previously developed an algorithm for locating boundaries in an image with sub-pixel resolution, as well as estimating boundary width and image intensity within the adjoining objects. The algorithm operates by finding the parameters of a cumulative Gaussian curve that best approximates an intensity profile taken across a boundary. If intensity is sampled along the image gradient across a boundary, it is reasonable to assume the profile approximates a finite portion of a cumulative Gaussian. Given that assumption, the first derivative of the profile should be the corresponding portion of a Gaussian, completely described by its mean, standard deviation, and amplitude. We present here a simple and rapid method to find those parameters, given that we only have a potentially skewed sample of the Gaussian. The parameters are approximated first for the finite sample, and then both ends of the Gaussian are extrapolated using the resulting parameters. New parameters are then calculated and the procedure is repeated. The optimization rapidly converges, yielding boundary location (mean) with sub-pixel accuracy as well boundary width (standard deviation). Integration then reproduces the cumulative Gaussian, and a least-squares fit is applied to estimate the constant of integration, from which intensity of the adjoining regions can be estimated.

Cumulative Gaussian Curve Fitter for Boundary Parameterization

Using the path framework we previously added to ITK, we implemented a novel algorithm for n-dimensional path optimization, which we call the ND Swath (NDS). NDS uses dynamic programming to globally optimize the placement of a path within an image, subject to several constraints and a user-supplied merit function. The NDS algorithm is presented in this paper along with a description of how it was implemented using ITK.

N-Dimensional Path Optimization: The Implementation of a Novel Algorithm in ITK

Dynamic magnetic resonance imaging (DCE-MRI) carried out with
contrast media such as Gd-chelate complex (Gd-DTPA) allows the
non-invasive assessment of microcirculatory characteristics of
malignant lesions. Quantitative estimation of lesion parameters
from the passage of the contrast media requires the use of
pharmacokinetic two-compartment model. The input to the model is
the time-intensity plot from a region of interest (ROI) covering
the lesion extent. The lengthy imaging process, elasticity of the
organs and patient movement result in complex deformations in the
subject requiring 3D motion correction for ROI alignment. This
paper presents results on applying the Thirion Demon's 3D elastic
matching procedure in the ITK framework on the two-compartment
lesion parameters. Registration, meanwhile involves interpolation
and smoothing operations thereby affecting the time-intensity
plots. We explore the trade-offs that arise between registration
and lesion parameter estimation. Experiments on synthesized and
real deformation are presented.

Impact of motion correction on the quantitative analysis of DCE-MR Images

The gold standard for measuring muscle regeneration in muscular dystrophy therapies is counting the number of dystrophin-positive muscle fibers on a cryostat muscle section immunostained for dystrophin. The standard process of manually counting a few thousand myofibers is tedious, time consuming, and limits quantitative analysis of a therapy success. We present an unsupervised method for segmenting and counting the number of myofibers on an immunofluorescent microscopy image. The key threshold selection problem is resolved by maximizing the number of sub-threshold connected components. Components significantly smaller than the known lower bound myofiber area, the only input parameter, are ignored to reduce noise. Validation on a series of images (n=63) revealed that our algorithm varied by less than 10% from manual counts in the relevant range of operation. The algorithm allows us to quantify three-dimensional dystrophin expression and design experiments that address a major limitation in muscular dystrophy therapies, the limited distribution of dystrophin after treatment. Further we have extended this method to segment and count objects in other immunofluorescent images. The method was quickly developed and tested using the Insight Toolkit (ITK), an open source C++ library for the development of image analysis software.

Unsupervised Segmentation for Myofiber Counting in Immunofluorescent Microscopy Images

We are developing open source software for computer assisted intervention systems. Our primary experience has been with medical
robots, but the concepts (and software) apply to many physical devices that interact with the real world.
The real-time performance requirements permeate all levels of our software, including common tools (such as logging, class and object registers),
vectors, matrices and transformations. Our software libraries are written in C++, but are also accessible from Python, which
provides a convenient environment for rapid prototyping and interactive testing.
The real-time support includes a device (hardware) interface and a task library. Device-specific modules such as robot servo control and trajectory generation can be provided by tasks or by external devices.
Ultimately, we intend to provide a framework that supports extension via dynamically loaded plug-in modules.
Our development process utilizes a multitude of open source tools, including CVS, CMake, Swig, CppUnit, Dart, CVSTrac, Doxygen
and LaTeX. These tools help to ensure compliance with our software development procedure.

Development of open source software for computer-assisted intervention systems

Salient points are used for various applications, such as medical image registration, tracking, stereoscopic matching. The purpose of this paper is to compare two commonly used methods to extract salient points in 3D medical images. We give an interpretation of the methods and validate their performance empirically based on criteria derived for the task of image registration, displacement measurement and tracking in medical images.