Quantitative CT sharpens IPMN risk prediction

In resected IPMNs, adding quantitative CT features to guideline criteria modestly improved prediction of high-grade dysplasia or cancer—potentially useful for risk assessment, but not yet ready to guide surgery.

• The combined model’s AUC was 0.800 versus 0.760 for guideline criteria alone (p=0.007).
• Pancreatic HU variability was higher in high-grade dysplasia and malignancy than in low-grade dysplasia; pancreatic index fell as disease grade increased.

This single-center retrospective study included only patients who underwent resection. External validation is needed before applying the model to broader IPMN populations or changing operative decisions.