Data Mining for Brain Imaging

Wednesday, March 3, 2004 - 17:30
TH 331
Aleksandar Lazarevic, University of Minnesota

Current advances in medical image acquisition techniques have made available enormous amounts of remarkable high-resolution three-dimensional (3D) image data. This talk focuses on several novel methods for classifying medical images by focusing on the regions of interest (ROIs). Two statistical methods for classifying ROIs in medical images are presented based on their three dimensional spatial distributions. The first statistical method is based on measures of dissimilarity between probability distributions. Measures such as the Mahalanobis distance and the Kullback-Leibler divergence between the spatial arrangement of ROIs of a new subject and those of previously seen data sets related to each considered class are computed. A new subject is predicted to belong to the class corresponding to the dataset that has the smallest distance from the given subject. The second statistical method is based on maximum likelihood employing expectation-maximization and clustering algorithms to estimate the underlying distributions. Finally, a novel Dynamic Recursive Partitioning (DRP) approach for discovering discriminative patterns of functional MRI activation will also be presented. All methods were experimentally evaluated on mixtures of Gaussian distributions, on realistic lesion-deficit data generated by a simulator conforming to a clinical study, and on functional MRI activation data obtained from a study designed to explore neuroanatomical correlates of semantic processing in Alzheimer's disease. Experimental results on all data sets have shown a great potential in facilitating retrieval of similar images and the elucidation of associations between image and other clinical (e.g., behavioral) data.


Aleksandar Lazarevic is a Research Associate at Army High Performance Computing Research Center, University of Minnesota . His research interests include data mining, parallel and distributed computing as well as intrusion detection. He received B.Sc. and M.Sc. degrees in Computer Science and Engineering from the University of Belgrade , Yugoslavia in 1994 and 1997 respectively. He received his PhD degree in Computer Science from Temple University in December 2001. He has authored around 30 research articles. Starting from January 2002, he is leading the project related to applications of data mining for network intrusion detection. He served as a Co-Chair for the Workshop on Data Ming for Cyber Threat Analysis at the IEEE International Conference on Data Mining, held in Japan in December 2002. He also served as Program Committee member at the same conference, at the Pacific Asia Conference on Knowledge Discovery and Data Mining 2003 and 2004, at the Seventh International Conference on Discovery Science, 2004 and at several workshops. He also served as a Publicity Chair for the SIAM International Conference on Data Mining 2003 and 2004. He is a member of IEEE , SIAM and ACM.