Classifyber

Classifyber segments white matter bundles from diffusion Magnetic Resonance Imaging (dMRI) by applying a supervised streamline-based linear classifier for delineation of white matter pathways.

Key Features:

• Supervised streamline-based segmentation: Operates on tractography-derived streamlines using supervised learning to assign streamlines to white matter bundles.
• Linear classifier: Implements a linear model that combines multiple feature sources into a single decision boundary for streamline classification.
• Integrated information sources: Leverages anatomical atlases, connectivity patterns, and geometric characteristics of fiber pathways as input features.
• Input data: Processes diffusion Magnetic Resonance Imaging (dMRI) data and tractography outputs.
• Robustness to methodological variation: Targets robustness across diverse tracking methodologies, bundle sizes, and data qualities.
• Segmentation performance: Demonstrates improved segmentation outcomes across multiple datasets ranging from research-focused to clinical applications.

Scientific Applications:

• Pre-surgical planning: Facilitates delineation of patient-specific white matter bundles to inform surgical approaches.
• Connectomics: Enables bundle-level analyses for connectomics studies using dMRI-derived tractography.
• White matter pathway delineation: Supports virtual delineation and comparative analysis of white matter pathways in research and clinical dMRI datasets.

Methodology:

Supervised streamline-based linear classification applied to tractography streamlines, integrating anatomical atlas labels, connectivity-derived features, and geometric descriptors into a linear model.