Table 1.

Radiomic models for diagnosis and risk stratification of IPMNs.

Study	Sample Size	Data	Best-Performing Model	Task	AUC	Comparisons
Permuth, 2016 [53]	38	CT texture analysis + genomics	Logistic regression	Distinguish malignant from benign IPMNs	0.92	N/A
Hanania, 2016 [54]	53	CT imaging (texture, shape, intensity)	Logistic regression	IPMN high- vs. low-grade dysplasia	0.96	Lower false positive rate than Fukuoka
Chakraborty, 2018 [55]	103	CT imaging features	Random forest	High- vs. low-risk BD-IPMN	0.77	N/A
Corral, 2019 [56]	139	MRI imaging features	CNN	Identify high-grade dysplasia or cancer in IPMNs	0.78	Accuracy was comparable to AGA/Fukuoka
Chu, 2022 [57]	214	CT radiomics features	Random forest	Classify mucinous and non-mucinous cysts	0.94	Accuracy was comparable to radiologist
Liang, 2022 [58]	193	CT + clinical data	Fused radiomics-DL	Differentiate MCN from IPMN	0.973	N/A