Computer vision algorithm outperforms experts in predicting SMA margin status.

A computer vision algorithm was developed to predict superior mesenteric artery (SMA) margin status in pancreatic ductal adenocarcinoma patients using preoperative CT scans. In a study involving 200 patients, the algorithm achieved a sensitivity of 0.43, outperforming radiologists (0.23) and surgeons (0.36). Specificity was high across the board, with both the algorithm and a radiologist achieving 0.94. Overall accuracy for the algorithm was 0.85, surpassing radiologists (0.80) and surgeons (0.76), indicating significant potential for clinical application.