Table 1.

AI’s role in pancreatic cancer.

Study	Sinkala M et al., 2020, South Africa [8]	Wazir M. et al., 2019, USA [15]	Corral JE et al., 2019, USA [16]
AI type	ANN	ANN	ANN
Topic	differentiation of molecular-genomic profile of PDAC subtypes (mRNA and DNA methylation models)	PDAC risk prediction based on personal health data	IPMN characterization on MRI
Study population	45 pancreatic cancer samples	898 patients diagnosed with pancreatic cancer	139 patients with histologically characterized IPMNs (due to pancreatectomy) and previous MRI images
Main results	AI in differentiation of two PDAC subtypes: overall classification Ac 100% for the mRNA-based model, 99% for the DNA methylation-model; model provides predictions of clinical response to chemotherapy	AI Sn and Sp in testing cohort: 80.7%, 80.7%; AUROC curve 0.85.	AI in detect dysplasia Sn and Sp: 92%, 52%. Identification of high-grade dysplasia/cancer: Sn and Sp 75% and 78%. AI AUROC curves 0.78 (p = 0.90) vsAUROC base on AGA criteria 0.76, AUROC based on Fukuoka criteria 0.77.

Abbreviations: AI (Artificial Intelligence), PDAC (Pancreatic Ductal Adenocarcinoma), ANN (Artificial Neural Network), mRNA (Messenger RNA), DNA (Deoxyribonucleic Acid), IPMN (Intraductal Papillary Mucinous Neoplasm), MRI (Magnetic Resonance Imaging), Ac (Accuracy), Sn (Sensitivity), Sp (Specificity), AUROC (Area Under the Receiver Operating Characteristic Curve), and AGA (American Gastroenterological Association).