C.N.S. Introduction Page

This page briefly presents a few screenshots of some aspects of the "Coregistration for Neuroimaging Systems" (C.N.S.) neuroimaging software package. C.N.S. was originally developed with Drs. Linda Chang and Thomas Ernst at Harbor-UCLA Medical Center, Torrance, California. We are using it and continuing its development at USC.

The software consists of over 300 processing modules and over 250 networks of such modules, written in AVS. It includes, among others, custom automated algorithms for skull stripping, surface-based coregistration, segmentation of CSF, correction for partial volume effects in SPECT/PET/pMRI, coregistration of MRS to MRI allowing accurate MRS localization on follow-up scans, MRI morphometry and drawing of regions of interest, extraction of white-matter lesions, correction for MRI intensity decay with surface coils, correction for geometric distortions and patient motion in EPI-fMRI, sparse fusion of fMRI time series across sessions, computation of fMRI activation, computation of blood flow in Gd-pMRI, elimination of large vessels in pMRI, computation of diffusion tensor in dMRI, 3D surface mesh reconstruction and optimization, and various 2D and 3D visualization tools, classical image processing tools and image conversion tools.

The system can be simultaneously built and executed by an unrestricted number of users on various hardware platforms, can automatically generate restricted distributions and updates from the master system, and automatically generates LaTeX and HTML manuals from the online help pages.

Brain Segmentation
The software includes tools for automated segmentation of the brain from surrounding membranes, skull and muscles. The segmentation programs use simple mathematical morphology operators to erode out non-brain objects.
Surface Coregistration
Once brains have been segmented from two scans, the resulting 3D surfaces can be re-oriented such as to match in position and 3D orientation. This process uses an anisotropic chamfer distance map to accelerate the computation of mismatch between the two surfaces, and Powell's iterative multidimentional optimization algorithm to perform the matching. Additional visual inspection and correction tools are provided as well.
Correction for Partial Volume Effects
After coregistration, high-resolution information extracted from MRI can be used to correct a posteriori for partial volume effects occuring in lower-resolution functional scans (such as PET and SPECT). The algorithn consists of segmenting from MRI all regions of the brain which are known not to exhibit any radiotracer uptake (CSF, bone, air), and of using this information to redistribute measured radioactivity only in the regions which are known to show tracer uptake (this process hence ressembles a de-convolution).
Display and Measurements
Tools are provided to display results and perform quantitative measurements, including the manual drawing of regions of interests on multiple coregistered scans from any modalities.
Other Processing
Other processing capabilities are being developed, including automatic scan prescription, automatic segmentation of white-matter lesions, and many more.

Copyright © 2000 by the University of Southern California, iLab and Prof. Laurent Itti