AI model enhances diagnosis of anal injuries

A novel convolutional neural network (CNN) algorithm for endoanal ultrasound (EUS) efficiently distinguishes between benign anal injuries, notably achieving 100% accuracy for fissures. In assessing external and internal lacerations, the model demonstrated 82.5% sensitivity and 93.5% specificity, alongside 91.7% sensitivity and 85.9% specificity, respectively. This groundbreaking AI-assisted approach could significantly enhance diagnostic accuracy in proctology, reducing reliance on expert interpretation and supporting wider clinical application despite its initial single-center limitations.