Version 2.5.2.0

Abstract

Grant Number:	5R01LM005944-03
Project Title:	ALGORITHMS FOR COMPRESSION AND REGISTRATION OF BRAIN MRI

PI Information:	Name	Email	Title
	SAHNI, SARTAJ K.	sahni@cise.ufl.edu	DISTINGUISHED PROFESSOR AND CHAIR

Abstract: Magnetic resonance imaging of the brain has made possible to test old hypotheses about the relation of cognition to human brain structure. A key requirement for reliable measurement is the ability to relate areas from scans taken at different time points, different devices, and different individuals. The goal of this project is to develop efficient algorithms for (a) compression of MR volume data of the human brain and (b) registration of MR volume data sets taken of a human brain before and after epilepsy surgery. These volume data occupy a large amount of space and thus can prove to be time intensive to access, manipulate and transmit over computer networks. Thus, the data should be represented in a compact form which would require far less space and would be easy to access as well as manipulate. In this proposal, we compare the performance of some existing standard, static and dynamic image compression algorithms applied independently to the MR data and then point out some of their pitfalls thus justifying the need for a proposed new volume data compression algorithm. The problem of registration is very important in accurately determining the location and amount of surgically removed tissue etc. In this project, we propose a novel technique for registering the pre and post operative volume data using the 3D shapes recovered from the brain MRI. A coarse to fine registration method is proposed in which gross anatomical structures in the neighborhood of the anatomical shape of interest are used as the shapes to be registered at a coarse scale and the shape of interest itself are registered at a fine scale. Registering gross features/shapes at a coarse resolution provides an initial estimate on the registration function which is modeled as an affine transformation (rotation, translation and scaling). This estimate can then be used as an initial guess for the registration function at a finer resolution for registering detail. We present preliminary results for MRI data compression and for 3D shape recovery from MR brain scans. These 3D shapes will be used as input to the proposed registration algorithm. The proposed algorithms for compression and registration of MRI data are not limited for use with MRI data but, may be used with other types of volume data as well. The results should improve the ability of neuroscientists, physicians, and cognitive scientists to assess the impact of such variables as development and surgery on brain region characteristics and function.

Public Health Relevance:
This Public Health Relevance is not available.

Thesaurus Terms:
brain scanning, computational neuroscience, image processing, magnetic resonance imaging
brain mapping, computer program /software, epilepsy, image enhancement, neurosurgery, postoperative state, preoperative state
data collection methodology /evaluation, human subject

Institution:	UNIVERSITY OF FLORIDA
	219 Grinter Hall
	GAINESVILLE, FL 326115500
Fiscal Year:	1997
Department:	COMPUTER AND INFORMATION SCIS
Project Start:	15-AUG-1995
Project End:	14-AUG-1999
ICD:	NATIONAL LIBRARY OF MEDICINE
IRG:	BLR

---