UBO Detector - A cluster-based, fully automated pipeline for extracting white matter hyperintensities

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• White matter hyperintensities (WMH) are common in aging and neurological diseases.
• Accurate segmentation of WMH is crucial for understanding disease progression and treatment effects.
• Current WMH segmentation methods can be time-consuming and subjective.

Purpose of the Study:

• To develop and validate 'UBO Detector', a fully automated pipeline for WMH segmentation.
• To provide an efficient and reliable tool for quantifying WMH volumes and clusters.
• To facilitate research on the clinical implications of WMH.

Main Methods:

• Input: T1-weighted and FLAIR MRI scans.
• Pre-processing: SPM12 and FSL.
• Segmentation: FMRIB's Automated Segmentation Tool (FAST) and k-nearest neighbor (k-NN) machine learning.
• Output: Image and text data on WMH volumes and cluster counts.

Main Results:

• High agreement with manual WMH tracing (SI=0.848, ICC=0.985).
• Accurate tracking of WMH volume changes over time (r² > 0.9).
• Significant correlation of WMH volumes with Fazekas scores (p < 0.001).
• Processing time: approximately 15 minutes per brain.
• Parallel computing enables efficient multi-core CPU utilization.

Conclusions:

• UBO Detector is a reliable, efficient, and fully automated pipeline for WMH segmentation.
• The tool provides accurate and reproducible WMH quantification.
• UBO Detector can significantly aid neuroimaging research and clinical applications involving WMH.