The main objective of this proposal is to develop a major software package for automatically (or with minimal manual intervention) processing a large number of magnetic resonance images that are typically acquired in multi- center clinical trials for estimating volumes of brain tissues. The software will be based on the following two modules: an image format library, and an image control environment. (l) The image format library (IFL) will provide an abstraction, allowing medical images from different magnetic resonance (MR) scanners to be manipulated using a common internal format that is simple and flexible, and allows access to all the necessary information contained in the files without tethering the processing code to a specific file or storage format. IFL will provide interfaces for reading, writing, creating, duplicating, and extracting study information from medical images, and will be structured such that adding new formats will be quite straightforward and transparent to the processing and viewing code. (2) The Image control environment (ICE) will use the IFL library to manage an internal catalog of images, and will orchestrate the sequence of operation on these images. In particular, it will incorporate a set of dimensional registration, image segmentation methods and volume measurements of individual tissues. ICE will also incorporate a specialized image viewer optimized for multiple medical image display. The ICE module will be structured such that any user defined functions can be incorporated easily without affecting the other components. This module will provide interfaces for manual and scripted image processing operation on series of images, and will provide advanced macro capabilities. These major software modules will be seamlessly integrated for the automated image processing. In phase 1, we will concentrate on developing IFL and ICE and integrating them. The individual modules and the integrated software package will also be evaluated in phase l using phantom and stimulated brain images. In phase 2 we will further refine this software, based on phantom studies, and evaluate it using brain images. At the end of the phase 2 studies, we expect that this software will be completely automated and tested on normal and patient brain images generated with a variety of pulse sequences on different MR scanners.
Analysis of the vast amount of MR data that is typically acquired in multi- center clinical trials requires image processing software that can automatically analyze and quantitate the image data. The proposed software is expected to fill this void, hence the commercial potential is significant.
