Deep Learning-Driven Prediction of Pediatric Spinal Cord Injury Severity Using Comprehensive Structural MRI Analysis

Motivation: Diagnosing pediatric spinal cord injury (SCI) is challenging due to limitations in current clinical assessments. Goal(s): Our study investigates how structural changes in the pediatric spinal cord can predict SCI severity, using cross-sectional measurements and deep learning for ASIA Impairment Scale classification. Approach: We analyzed spinal MRI data from 61 pediatric subjects, measuring cross-sectional area, anterior-posterior, and right-left widths across vertebral levels. A convolutional neural network was trained on these structural features, combined with age and height. Results: The model achieved 96.6% accuracy in classifying SCI and 94.9% accuracy in predicting ASIA categories, identifying significant structural differences between SCI and TD groups. Impact: This research identifies structural MRI biomarkers for pediatric SCI severity, offering a precise tool for assessing injury severity. The approach offers clinicians a potential tool for refined injury assessment and sets a foundation for further advancements in pediatric SCI management.