AUTOMATIC SULCAL CURVE EXTRACTION WITH MRF BASED SHAPE PRIOR

Extracting and labeling sulcal curves on the human cerebral cortex is important for many neuroscience studies, however manually annotating the sulcal curves is a time-consuming task. In this paper, we present an automatic sulcal curve ex-traction method by registering a set of dense landmark points representing the sulcal curves to the subject cortical surface. A Markov random field is used to model the prior distribu-tion of these landmark points, with short edges in the graph preserving the curve structure and long edges modeling the global context of the curves. Our approach is validated using a leave-one-out strategy of training and evaluation on fifteen cortical surfaces, and a quantitative error analysis on the ex-tracted major sulcal curves. Index Terms — Sulcal curve extraction, cortical surface, Markov random field, shape prior, point set registration 1