GROUP-WISE CONSISTENT SULCAL FUNDI SEGMENTATION BASED ON DMRI-DERIVED ODF FEATURES

Recently, it has been shown that structural connectivity patterns derived from diffusion MRI (dMRI) can be used for cortical parcellation and segmentation. However, most previous methods were based on diffusion tractography, which is limited in depicting local profiles, e.g., in regions beneath the cortical sulcal fundi. Instead, in this paper, we propose to derive effective features directly from orientation distribution functions (ODFs), which contain rich information about local fiber connection profiles like crossings. Based on these ODFs features, we adopted a multi-view spectral clustering method to group-wisely and consistently parcellate the cortical sulcal fundi. Experimental results on the central sulcal fundi are promising, which is further validated by resting state fMRI data mapping.