Table 1.

The main characteristics of the included studies in the meta-analysis.

No. crt.	Author	Study Design	Comparison	No. of Patients (Overall Data)	No. of Images (Overall Data)	Testing Data	Final Diagnosis	Analysis Target	Type of Computer-Aided Diagnosis (CAD)	Algorithm of AI
1	Kuwahara 2019	retrospective	benign IPMN vs malignant IPMN	50	3970	no separate testing data	27 benign IPMN/23 malignant IPMN	B-mode images	deep learning-based CAD	CNN
2	Das 2008	retrospective	normal pancreas vs. chronic pancreatitis (CP) vs PDAC	56	319	50% of all data	2 normal pancreas/12 CP/22 PDAC	Texture features from B-mode image	conventional CAD	ANN
3	Marya 2020	retrospective	autoimmune pancreatitis vs. normal pancreas vs. CP vs. PDAC	583	1,174,461	123 patients	146 AIP/292 PDAC/72 CP/73NP	B-mode images	conventional CAD	CNN
4	Norton 2001	retrospective	CP vs PDAC	35	N/A	N/A	14 CP/21 PDAC	Grey-scale pixels from B-mode image	conventional CAD	Basic Neuronal Network/Machine Learning
5	Ozkan 2015	retrospective	PDAC vs. normal pancreas	172	332	72 (42 PDAC, 30 normal pancreas)	202 PDAC/130 normal pancreas (images)	Digital features from B-mode image	conventional CAD	ANN
6	Saftoiu 2015	Prospective	CP vs. PDAC	167		15% of pts	112 PDAC/55 CP	TIC parameters from contrast-enhanced EUS	conventional CAD	ANN
7	Tonozuka 2021	Prospective	normal pancreas vs. CP vs. PDAC	139	1390	47 pts, 470 images (25 PDAC, 12 CP, 10 NP)	76 PDAC/34 CP/29 normal pancreas	B-mode images	deep learning-based CAD	CNN
8	Udristoiu 2021	Retrospective	CP vs. PDAC vs. NET	65	3360	672 images from 65 pts	30 PDAC 20 CP/15 NET	Multi parametric (B-mode, contrast, elastography)	deep learning-based CAD	CNN
9	Zhang 2010	Retrospective	CP vs. PDAC vs. normal pancreas	216		50% of all data	153 PDAC/20 normal pancreas/43 CP	Texture features from B-mode image	conventional CAD	SVM
10	Zhu 2013	Retrospective	CP vs. PDAC	388		50% of all data (194; 131 PDAC, 63 CP)	262 PDAC/126 CP	Texture features from B-mode image	conventional CAD	SVM