A. Ekin; "Computer-Aided Brain MRI analysis..." 20.12.2005, 11:40,G035

Faculty of Engineering and Natural Sciences
FENS SEMINARS

Computer-Aided Brain MRI Analysis for Early Detection of Neurodegenerative Diseases

Ahmet Ekin
Senior Scientist, Video Processing Group, Philips Research Europe
High Tech Campus 36, Eindhoven, The Netherlands

Abstract
Most neurodegenerative diseases, including Alzheimer’s and Parkinson’s, are associated with a large amount of iron accumulation in the basal ganglia region of the brain. Because of this, it is often claimed that iron could be a potential biomarker for such diseases. Early detection of these diseases by iron accumulation features is very appealing because brain tissues with high iron concentration appear darker in the T2 contrast of MR images and can be detected by image processing methods without any invasive procedure. In this talk, I will describe our work at Philips Research on brain MR image analysis to automatically detect the iron accumulation in the basal ganglia region of the brain. After brief introductory information about Philips and the general activities in the Video Processing group, I will focus on the first four steps of the fully-automated MR-based brain iron detection system: 1) fast feature point based head orientation estimation to register head to the standard coordinates, 2) clustering-based brain tissue segmentation to identify cerebrospinal fluid region, 3) shape-based volume-of-interest detection to automatically locate the basal ganglia region, and 4) contrast-robust tissue segmentation to identify white matter and gray matter regions. The proposed algorithms have been validated with MR brain scans of 50 elderly patients and have shown excellent performance.   

Short bio of Ahmet Ekin
Ahmet Ekin is a senior scientist in the video processing group at Philips Research Netherlands. He received his MS and PhD degrees from the Electrical and Computer Engineering, University of Rochester, Rochester, NY, USA, in 2001 and 2003, respectively, and his BS degree with the highest honors from the Electrical and Electronics Engineering, Bogazici University in 1999. In Philips, he has been first involved in video processing for consumer electronics, and then, computer vision for night vision applications, and recently, medical image analysis. He co-initiated the medical image processing activity in the video processing group that resulted in 600K- and 1.5mio-euro funding for 2005 and 2006, respectively. His pedestrian detection algorithm for night vision achieved the best result in the BMW Concept competition in Jan. 2005. His doctoral study on sports video summarization has been one of the initial and most-cited works in the sports video processing area. During his doctoral years, he also worked at Eastman Kodak Co., Rochester, NY as a consultant on video modeling for the MPEG-7 standard, spent the summer of 2001 at AT&T Labs, Middletown, NJ working on content-based video coding, and co-designed the Face Cataloger of IBM PeopleVision video surveillance system between June-Nov. 2003 as a technical co-op at IBM T.J. Watson Research Center. He has written more than 25 papers or book chapters, and has 11 pending U.S./E.U. patents in a wide range of areas, including statistical object detection, color image processing for digital cameras, video surveillance, and medical image processing and visualization.     

December 20, 2005, 11:40, G035