Meeting Banner
Abstract #1924

Automated Hierarchical Clustering of DTI White Matter Fiber Tracts

Zhenyu Zhou1, Yijun Liu2, Guang Cao1, Karen M. von Deneen2, Dongrong Xu3

1Global Applied Science Laboratory, GE Healthcare, Beijing, China; 2McKnight Brain Institute, University of Florida, Gainesville, FL, United States; 3MRI Unit, Columbia University, New York, NY, United States

Recently, the tract-based analysis of white matter fibers has raised interests from the neurology and clinical neuroscience community since this methodology provides quantitative analysis of the properties of the specific fiber bundles, which provide a useful abstraction of the white matter structures and a clear identification of neural fibers. In order to benefit from the tract-based analysis, many clustering algorithms of the fiber tracts have been proposed. However, most approaches require a user initialization. In this paper, we propose a novel cluster method to automatically group brain white matter fibers into biologically meaningful neural tracts.

Abstract PDF

Keywords

ability abstraction achieves allows anatomical applicability applied approaches atlas automated automatically benefit biologically brain bundle bundles characterizing china classification clear clinical cluster clustering collected colored commonly community computational conjunction connectivity corpus correction currently damage demand depict detecting developed diffusion distance eddy effective eigenvector elements ensemble ensuring every expansion experimental features fiber fibers field fields flowchart framework full furthermore future fuzzy generated geometry global good growing hierarchical human identification implements include inner institute interacting invoked isotropic labeled laboratory look many material meaningful methodology neural neurology normalized novel paper part people performance population positive potential predicting preprocessing principal prior procedures project promising properties propose proposed prospect psychiatry quantitative raised rapid reasonably recently reduces republic require resolution robust scanner seeding segment segmentation semi several since sites space spatially spinal squared steps structure structures studies succeed system task tensor tissue tissues together tool tracking tract tracts tree underlying unit unsupervised useful user white whole yielding