Automatic brain quantification in children with unilateral cerebral palsy

Area of Science:

• Neuroimaging
• Pediatric Neurology
• Artificial Intelligence in Medicine

Background:

• Assessing brain damage in children with spastic unilateral cerebral palsy (uCP) presents significant clinical challenges.
• Accurate quantification of brain structures and lesions is crucial for understanding disease progression and outcomes.

Purpose of the Study:

• To develop and validate a deep learning pipeline for automated quantification of lesion-free brain volumes in children with uCP.
• To correlate automated volume measurements with lesion extent and clinical outcomes.

Main Methods:

• Utilized T1-weighted and FLAIR MRI data from pediatric patients.
• Developed deep learning models for automatic brain structure and lesion segmentation.
• Validated the pipeline on independent datasets and clinical evaluations.

Main Results:

• The models demonstrated robust segmentation performance, even in cases with severe brain alterations.
• Identified reduced lesion-free volumes in the affected hemisphere, correlating with lesion extent (p < 0.05).
• Found associations between regional lesion-free volumes (e.g., thalamus) and motor/visual outcomes (p < 0.05).

Conclusions:

• Automated lesion-free volume quantification is a valuable tool for assessing brain damage in uCP.
• This approach aids in exploring brain structure-function relationships in pediatric neurological disorders.
• Supports the clinical utility of AI-driven neuroimaging analysis for improved patient care.