Study Population and Image Analysis

The present multicenter retrospective study included 120 patients with pathologically confirmed pancreatic IPMN who underwent CT between 2010 and 2023. Eight reviewers – four abdominal radiologists and four radiology residents – interpreted the CT images, and diagnostic performance and interobserver agreement were analyzed using Fleiss κ statistics and receiver operating characteristic analysis, based on both the IAP guidelines and the reviewers’ experience.

For each of the 120 patients, the radiologists evaluated the following parameters related to IPMNs (5–7,10–12,15–19): (a) IPMN type (main, branch, or mixed duct type); (b) presence of enhancing mural nodule(s), if any, and size of the largest (mm); (c) size of the main pancreatic duct (MPD) (mm); (d) presence of cystic lesion(s), if any, and size of the largest (mm); (e) assessment of thickened or enhancing cyst walls; (f) identification of abrupt MPD caliber changes with distal pancreatic atrophy; (g) presence of lymphadenopathy; (h) pancreatitis; and (i) infiltrative masses (solid masses extending beyond the margin of the cyst or dilated MPD). Following the 2017 IAP guidelines (7), we documented high-risk stigmata and worrisome features. Ancillary high-risk stigmata, as described by Kang et al., included the categorization of infiltrative masses.

Assessment of Malignant Potential in Pancreatic IPMNs: Diagnostic Proficiency According to the IAP Guidelines

After the radiologists recorded the presence or absence of high-risk stigmata and worrisome features, the presumed malignant potential was graded using a five-point scale (Figure 2), based on the number of high-risk stigmata or worrisome features, as follows: a score of 5 indicated at least one high-risk stigmata; a score of 4 indicated any three or more worrisome features; a score of 3 indicated any two worrisome features; a score of 2 indicated any worrisome features; and a score of 1 indicated no worrisome features or high-risk stigmata.

Two additional binary scales were used. On binary scale I, a score of 5 was categorized as malignant, whereas scores of 1–4 were considered benign, and on binary scale II, a score of 4 or 5 was designated as malignant, and scores of 1–3 were classified as benign.

Assessment of Malignant Potential in Pancreatic IPMNs: Diagnostic Proficiency Based on the Reviewers’ Experience

Following a washout period of four weeks, all reviewers independently reviewed the same complete set of anonymized CT images in a different randomized order. The clinical information was the same as that in the initial reading session. They recorded their confidence level concerning the malignant potential of pancreatic IPMNs using a five-point Likert scale, relying on their intuition based on accumulated experiences as follows: a score of 1 was definitely benign; a score of 2 was probably benign; a score of 3 was indeterminate; a score of 4 was probably malignant; and a score of 5 was definitely malignant.  Additional binary scales were also applied. For binary scale I, a Likert scale score of 4 or 5 was considered malignant, while a score of 1–3 was considered benign, and for binary scale II, a Likert scale score of 3–5 was classified as malignant, while a score of 1 or 2 was classified as benign.