千葉大学国際医工学セミナー

International Seminar Series on Biomedical Engineering

第14回
日時：2012年3月30日（金）13:00 - 14:00
場所：千葉大学ＣＦＭＥ　Ｂ号棟１階会議室　（西千葉キャンパス）

TITLE

the intersection of Machine Learning, Medical Imaging, and Neuroscience

LECTURER

Chris Hinrichs

(Ph.D. student at the University of Wisconsin in Computer Sciences)

ABSTRACT

This talk is including several projects:
Classification of Alzheimer's disease in MRI and FDG-PET scans, using a modified LP Boosting framework which is designed to work specifically in medical images, as well as in a multikernel framework. I also adapted a Multi-Kernel Learning (MKL) model to explicitly include outlier detection, and applied it to the same data. An offshoot of my work with Alzheimer's Disease data is an investigation into possible outliers in the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. As of this writing, the anomalies we are picking up could be due either to problems with the Voxel Based Morphometry (VBM) pre-processing pipeline, or the subjects themselves could be misdiagnosed. I am currently investigating other approaches to resolve this issue. I've also been involved in adapting these methods to work with axon bundle tracts which are segmented from Diffusion Tensor Images (DTI) on the problem of classifying Autism.