Chen, J.; Shi, D.; Su, J.; Huang, X.; Qian, Y.

The severity stratification of carpal tunnel syndrome (CTS) relies on ultrasound morphological markers and electromyography. However, it remains unclear how structural imaging can reliably infer functional impairment. Clarifying the structure-function relationship is critical for efficient diagnostic pathways. A retrospective cohort of 55 patients with symptoms related to CTS was analyzed at the Shanghai Sixth People's Hospital. All patients were subjected to ultrasound and EMG. 72.7% cases were diagnosed with CTS with a female predominance and equal left-right involvement. Random-forest classifiers were trained using surrogate splits, and performance was evaluated using predictions outside the bag. A full-feature model (34 candidate variables) was compared against a simplified model (8 core variables) capturing the core morphological and electrophysiological features. A residual-based framework was then used to characterize the structure-function mismatch within severity grades (1a-3c). The simplified model improved discriminative performance compared to the full-feature model (AUC 0.789 to 0.824). The simplified model achieved an overall accuracy of 77.3%. Analysis of predicted probability distributions and 10-bin calibration curves indicated stable and clinically interpretable risk estimation in most probability ranges. Permutation-based importance analysis confirmed that both ultrasound and electrophysiological features contributed substantively to prediction. Residual-based grading further revealed structure-function heterogeneity within each main severity grade. CTS severity can be stratified using a limited set of complementary morphological and electrophysiological features. Structure-function mismatch supports an imaging-led initial screening, with electrophysiology reserved for selected patients.