Machine learning in endoscopic ultrasonography and the pancreas: The new frontier?

Table 1 Summary of included machine learning studies on endoscopic ultrasonography in pancreatic disease

Field	Ref.	Study population used for training (n)	Task	Machine learning method	Performance (in test population if available)
Pancreatic Cysts	Kuwahara et al[23], 2019	Benign IPMN (27); Malignant IPMN (23)	Differentiate benign from malignant IPMN	Convolutional neural network	AUC = 0.98
	Springer et al[20], 2019	Mucinous cystic neoplasms (153); Serous Cystic Neoplasms (148); IPMN (447); Malignant cysts (114)	Guide clinical management by classify into three risk groups: No risk of malignancyLow risk of progression. High-risk of progression or malignant	Not available	First group: 100% specificity, 46% sensitivity. Second group: 54% specificity, 91% sensitivity. Third group: 30% specificity, 99% sensitivity.
	Kurita et al[22], 2019	Mucinous cystic neoplasms (23); Serous Cystic Neoplasms (15); IPMN (30); Other cyst types (17)	Differentiate benign from malignant cyst	Multi-layered perceptron	AUC = 0.96, sensitivity: 95%, specificity: 91.9%
	Nguon et al[21], 2021	Mucinous cystic neoplasms (59); Serous Cystic Neoplasms (49)	Differentiate mucinous cystic neoplasm and serous cystadenoma	Convolutional neural network	AUC = 0.88
Pancreatic Cancer	Saftouiu et al[27], 2008	PDAC (32); Normal pancreas (22); Chronic pancreatitis (11); Pancreatic neuroendocrine tumor (3)	Differentiate benign from malignant masses	Multi-layered perceptron	AUC = 0.96
	Saftoiu et al[28], 2012	PDAC (211); Chronic pancreatitis (47)	Differentiate cancer from benign masses	Multi-layered perceptron	AUC = 0.94
	Ozkan et al[30], 2016	PDAC (202); Normal pancreas (130)	Differentiate cancer from normal pancreas	Multi-layered perceptron	Accuracy: 87.5%, sensitivity: 83.3%, and specificity: 93.3%
	Udristou et al[31], 2021	PDAC (30); Chronic pancreatitis (20); Pancreatic neuroendocrine tumor (15)	Diagnose focal pancreatic mass	Convolutional neural network and long short-term memory	Mean AUC = 0.98 (Includes PDAC, CP and PNET)
	Tonozuka et al[32], 2021	PDAC (76); Chronic pancreatitis (34); Control (29)	Differentiate pancreatic cancer from chronic pancreatitis and normal pancreas	Convolutional neural network and pseudo-colored heatmap	AUC = 0.94
Autoimmune pancreatitis	Zhu et al[34], 2015	AIP (81); Chronic pancreatitis (100)	Differentiate AIP from chronic pancreatitis	Support Vector Machine	Accuracy: 89.3%, sensitivity: 84.1%, and specificity: 92.5%
	Marya et al[36], 2021	AIP (146); PDAC (292); Chronic pancreatitis (72); Normal pancreas (73)	Differentiate of AIP from PDAC	Convolutional neural network and pseudo-colored heatmap	AUC for AIP from all other = 0.92
Procedural assistance	Iwasa et al[38], 2021	Pancreatic mass (100)	Segmentation of pancreatic masses	Convolutional neural network	Intersection over unit = 0.77
	Zhang et al[40], 2020	EUS videos (339)	Recognition of stations, and segmentation of anatomical landmarks	Convolutional neural network	Accuracy for classification of stations (average) = 0.824, Dice coefficient for segmentation of pancreas (average) = 0.715

AUC: Area under the rule operator characteristic; AIP: Autoimmune pancreatitis; CP: Chronic pancreatitis; EUS: Endoscopic ultrasonography; IPMN: Intraductal papillary mucinous neoplasms; PDAC: Pancreatic ductal adenocarcinoma; PNET: Primitive neuroectodermal tumors.