Applications of Artificial Intelligence in the Diagnosis of Carpal Tunnel Syndrome: A Systematic Review

Introduction: Carpal tunnel syndrome (CTS) is an opportunity for artificial intelligence (AI) to be integrated into diagnostic practice. This study aimed to highlight AI technology being applied to the diagnosis, detection, and management of CTS.

Methods: A database search (PubMed, Embase, Cochrane) was conducted to identify any article associating “carpal tunnel syndrome” (and related Mesh terms) with “artificial intelligence” (and related Mesh terms). A total of 45 studies were identified that specifically applied AI to the diagnosis or detection of CTS. The pertinent findings of these studies were partitioned into diagnostic modality and highlighted in this review. When available, measures of accuracy (F1 score and AUC) were compared between AI frameworks.

Results: AI programs are capable of extracting radiomic data from a diagnostic image of a patient’s carpal tunnel with marked efficiency. Algorithms have been created to supplement ultrasound, MRI, nerve conduction studies, and thermographs. The enhanced detection and segmentation of the median nerve (MN) by AI technologies lends to its superiority to radiologist performance with respect to sensitivity and specificity of CTS diagnosis on ultrasound. Deep learning models are distinct from machine learning models, and their algorithm’s refinement had greater accuracy of CTS diagnosis when applied to US images, demonstrated by F1 score (p=0.01). Radiomic features of interest serving as inputs to AI programs include echointensity, pixelation patterns, and cross-sectional area of the MN. The highest degrees of diagnostic accuracy were achieved when excluding anatomic variants, including bifid MN and persistent median artery. AI correctly screened for CTS diagnosis and encouraged medical treatment from conversational text inputs. Patient surveys subject to AI analysis provided accurate prognostic outcomes, providing clinical data for consideration.

Conclusion : AI technology applied to CTS is capable of standardizing image acquisition and may eliminate discrepancies secondary to subjectivity. Data from patient questionnaires subject to AI analysis accurately predicted prognosis and may aid physician's determination of surgical candidacy. There is reason to investigate the integration of AI into clinical practice.