DeepnCCA

DeepnCCA performs automated segmentation of the corpus callosum from conventional MRI using deep learning to provide a neurodegenerative marker in multiple sclerosis (MS).

Key Features:

• Deep learning segmentation: Uses supervised machine learning algorithms to segment the corpus callosum from conventional MRI scans.
• Automated processing: Automates corpus callosum segmentation to replace manual 2D or volumetric 3D-based segmentations.
• High accuracy: Reports Dice coefficients of 98.1 ± 0.11% for intracranial area and 89.3 ± 0.76% for corpus callosum area, validated with 10-fold cross-validation.
• Clinical correlation: Outputs strongly correlate with clinical disability measures, including the Expanded Disability Status Scale (EDSS) ±6 months and future EDSS scores, and with future symbol digit modalities test results, showing numerically stronger correlations than FreeSurfer.
• Morphological analysis: Employs principal component analysis on segmentation outputs to characterize corpus callosum morphology, identifying thinning with increasing cognitive and physical disability and a more angled corpus callosum with increased physical disability.

Scientific Applications:

• Disease Monitoring: Correlates morphological changes of the corpus callosum with clinical disability measures to monitor MS progression over time.
• Therapeutic Response Evaluation: Provides quantitative corpus callosum measures that can be used to evaluate the effectiveness of therapeutic interventions in MS patients.
• Large Cohort Studies: Enables analysis of corpus callosum morphology across large MS cohorts for population-level studies of disease burden and progression.

Methodology:

Segmentation via deep learning using supervised machine learning on conventional MRI, validation with 10-fold cross-validation and Dice coefficients, and morphological characterization using principal component analysis with clinical correlation to EDSS and symbol digit modalities test results.