Visual interactive exploration and clustering of brain fiber tracts

Abstract

Nuclear magnetic resonance images have been used for detecting the movement of water molecules in living organisms and moreover exploiting the neural fibers distribution, which is of great significance for the brain disease analysis. However, due to the visual clutter of dense fiber tracts, it is difficult for medical researchers to understand the water molecule diffusion in whole-brain scale and to find the meaningful substructure of neurological pathways. To address the challenges, we provide one fiber visualization workflow that combines fiber tracts selection and fiber clustering approaches with the advanced visualization technique. Local and global fiber selection methods are provided for users to extract fiber tracts with the higher strength of water molecule diffusivity and gain an overall perception of water molecule movement in whole-brain scale. To explore the substructure of brain fibers, we employ an anatomically meaningful similarity matrix combining with density peaks clustering algorithm and compare it with DBSCAN algorithm. The qualitative and quantitative experimental results show that the fiber visualization system helps to confirm the fiber distribution more accurately and efficiently. .