Department of Biostatistics Seminar/Workshop Series
Statistical Methods for Label Fusion: Robust Multi-Atlas Segmentation
Bennett Landman, Ph.D
Assistant Professor, Department of Electrical Engineering, Vanderbilt University
Wednesday, March 20, 2013, 1:30-2:30pm, MRBIII Room 1220
Mapping individual variation of head and neck anatomy is essential for radiotherapy and surgical interventions. Precise localization of affected structures enables effective treatment while minimizing impact to vulnerable systems. Modern image processing techniques enable one to establish point-wise correspondence between scans of different patients using non-rigid registration, and, in theory, allow for extremely rapid labeling of medical images via label transfer (i.e., copying of labels from “atlas” patients to “target” patients). To compensate for algorithmic and anatomical mismatch, the state of the art for atlas-based segmentation is to use a multi-atlas approach in which multiple canonical patients (with labels) are registered to a target patient (without labels); statistical label fusion is used to resolve conflicts and assign labels to the target. We will discuss our recent progress (and outstanding challenges) in determining how to optimally fuse information in the form of spatial labels.