Classifyber, a robust streamline-based linear classifier for white matter bundle segmentation.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Medical Image Analysis

Background:

• Accurate white matter bundle segmentation is crucial for brain research and clinical applications.
• Current automated methods using diffusion MRI (dMRI) show limitations in quality and robustness across diverse datasets.
• Existing techniques rely on connectivity or fiber path geometry, with varying success.

Purpose of the Study:

• To introduce Classifyber, a novel supervised streamline-based method for segmenting white matter bundles.
• To improve the accuracy and robustness of white matter bundle segmentation compared to existing state-of-the-art approaches.
• To provide a versatile tool applicable to research and clinical dMRI datasets.

Main Methods:

• Classifyber employs a supervised, streamline-based approach integrating atlas information, connectivity patterns, and fiber path geometry.
• A simple linear model is utilized to combine these diverse data features for segmentation.
• The method was evaluated on multiple datasets spanning research and clinical domains.

Main Results:

• Classifyber demonstrated substantial improvements in segmentation quality compared to current state-of-the-art methods.
• The method proved robust across datasets with varying characteristics, including different tracking methods, bundle sizes, and data quality.
• Experimental results confirm enhanced performance in both research and clinical settings.

Conclusions:

• Classifyber offers a significant advancement in automated white matter bundle segmentation from dMRI data.
• Its robustness and improved accuracy make it a valuable tool for pre-surgical planning and connectomics.
• The open-source availability of Classifyber facilitates broader adoption and further research in neuroimaging.