Magnetic Resonance Image Processing

With the growth of 3D MRI, and the popularity of 4D functional MRI, there is a considerable need for advanced post-processing software to handle the extremely large volumes of data and to extract meaningful analyses. The primary analysis of such information is visualisation-the images are viewed by experts who can often judge immediately the significance of certain complex relationships. With very large datasets three problems arise-firstly the optimum method of visualisation is difficult to ascertain, secondly there is an inherent observer dependency, leading to problems in reproducibility and inter-observer variation, and thirdly there is a large demand on expert time. To deal with the first and third problems together, it is desirable to extract quantitative information such as volumetric measurements, and shape descriptions. This information is usually obtained manually by analysis of the images at a computer terminal. Segmentation that partitions the image data into anatomical compartments is the linking step between data acquisition, and quantitative measurement : Acquisition-> Segmentation-> Quantitative Measure A collaboration between University College London and the Institute of Neurology has been established for about five years on the automation of MR brain image segmentation. The original focus was on analysis of Multiple Sclerosis (MS) lesions, but we have now expanded to include applications in Epilepsy, and Psychosis.