ARTIFICIAL INTELLIGENCE AS A TRANSFORMATIVE FACTOR IN MOTILITY DISORDERS – AUTOMATIC DETECTION OF MOTILITY PATTERNS IN HIGH-RESOLUTION ANORECTAL MANOMETRY

Introduction Anorectal functional disorders are prevalent in clinical practice, presenting significant diagnostic challenges. High-resolution anorectal manometry (HR-ARM) is the first-line examination for assessing these disorders, but its application is limited by accessibility and the complexity of data interpretation. The aim of this study was to develop and validate an Artificial Intelligence (AI) model to identify and differentiate tone and contractility disorders in HR-ARM.

Methods A retrospective, single-center study was conducted using 813 HR-ARM exams. The classification of anorectal motility disorders was based on the London Classification. Several machine learning (ML) models were evaluated. The dataset was split, with 80% used for training and 20% for testing. The performance of the models was evaluated using accuracy, sensitivity, specificity, and positive and negative predictive values.

Results The LightGBM and xGB models demonstrated the highest overall accuracy. These models identified tone and contractility disorders with an accuracy of 97.6%, sensitivity of 98.9%, specificity of 94.7%, and positive and negative predictive values of 95.6% and 97.5%, respectively.

Conclusion The LightGBM and xGB models successfully identified and differentiated tone and contractility disorders in HR-ARM studies. This is the first global study to demonstrate the accuracy of ML models in detecting and differentiating motility patterns in HR-ARM. The development of AI models is crucial to improve the availability and accuracy of these exams, leading to more objective and precise management of patients with anorectal functional disorders.

 

MissingLinkText

No conflict of interest has been declared by the author(s).

Publication History

Article published online:
25 April 2025