Deep learning model detects acute pancreatitis on abdominal CT

A deep learning model detected acute pancreatitis on abdominal CT with an AUROC of 0.89 (internal) and 0.99 (external) in a retrospective multicenter study of 552 patients. External validation F1 score was 0.92.

• In internal testing (116 patients), the model achieved AUROC 0.89 (95% CI 0.82-0.95) and F1 score 0.83 (0.75-0.89).
• On independent external validation (95 patients), performance was robust: AUROC 0.99 (0.96-1.00) and F1 0.92 (0.86-0.97).
• The optimal model used both arterial and portal venous phase images; prospective validation is still required.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.