Diagnostic value of median nerve cross-sectional area measured by ultrasonography for the severity of carpal tunnel syndrome: a machine learning-based approach

Abstract

Objective: This study was conducted to evaluate the diagnostic performance and to establish cutoff values of median nerve cross-sectional area for classifying the severity of carpal tunnel syndrome.

Design: The study dataset included 1069 wrists from 1034 patients with carpal tunnel syndrome (May 2017 to December 2022). A machine learning algorithm was used to predict carpal tunnel syndrome severity based on median nerve cross-sectional area, adjusting for sex, age, body mass index, and disease duration.

Results: The multivariable model showed a multi-class AUC of 0.753, and single-class AUCs of 0.733, 0.635, and 0.780 for mild, moderate, and severe syndrome, respectively. Optimal cross-sectional area cutoffs were identified as <14 mm2 for mild and > 16 mm2 for severe syndrome, with AUC values of 0.773 and 0.794, respectively. The model showed high sensitivity for mild and high specificity for severe syndrome but had a low performance for moderate carpal tunnel syndrome (AUC = 0.568).

Conclusion: Median nerve cross-sectional area is a valuable tool for diagnosing mild and severe carpal tunnel syndrome. While cross-sectional area provides limited accuracy for moderate carpal tunnel syndrome, it remains a useful adjunct to other diagnostic methods, potentially reducing the need for more invasive procedures.