Random walks with efficient search and contextually adapted image similarity for deformable registration.

摘要: We develop a random walk-based image registration method [1] that incorporates two novelties: 1) progressive optimization scheme conducts the solution search efficiently via novel use of information derived from obtained probabilistic solution, and 2) data-likelihood re-weighting step contextually performs feature selection in spatially adaptive manner so data costs are based primarily on trusted sources. Synthetic experiments three public datasets different anatomical regions modalities showed our performed efficient without sacrificing accuracy. Experiments 60 real brain pairs dataset also demonstrated method’s better performance over existing non-probabilistic methods.