A Data-driven Voxel-wise White Matter Fiber Clustering Model Based on Priori Anatomical Data

Whole-brain fiber imaging allows nondestructive detection of human brain structural connections. The clinical application of this method is often classified as a series of fiber bundle structures of certain significance (function, structure, shape, etc.). Due to the lack of edge structure information of fiber bundles and the high variability of complex white matter structures in individual samples, fiber clustering based on anatomical information is still an open problem. In this paper, a new fiber clustering technique is proposed, which combines spatial features of whole-brain fibers and prior anatomical information as fiber similarity matching and feature extraction. In this work, we focus on the coverage of highly consistent fiber bundles in white matter structures to match anatomic features. The method is based on multiple tests of simulated data and in vivol data. The experimental results show that this method not only improves the highly consistent coverage of fiber bundles and prior anatomical knowledge, but also simplifies the fiber data space to improve the fiber clustering similarity measured population consistency. Finally, we also discuss the application of this method in clinical research.