Artificial intelligence: Emerging player in the diagnosis and treatment of digestive disease

Table 1 Recent researches on artificial intelligence in malignant digestive system diseases

Ref.	Study design	Study population	Year	Disease	Country/Region	Number of cases	Methods	Results
Tang et al[48]	Retrospective	Hospital	2020	GC	China	45240 Images	DCNN	Accuracy: 85.1%-91.2%; Sensitivity: 85.9%-95.5%; Specificity: 81.7%-90.3%; AUC: 0.887-0.940
Zhang et al[49]	Retrospective	Hospital	2020	GC	China	21,217 Images	CNN	Accuracy: 78.7%
Nagao et al[50]	Retrospective	Hospital	2020	GC	Japan	16557 images	CNN	Accuracy: WLI (94.5%); Accuracy: NBI (94.3%); Accuracy: Indigo (95.5%)
Song et al[51]	Retrospective	Hospital	2020	GC	China	3212 Images	DL	Accuracy: 0.873; Sensitivity: 0.996; Specificity: 0.843; AUC: 0.986
Ma et al[52]	Retrospective	Hospital	2020	GC	China	763 Images	CNN	Accuracy: 98.4%; Specificity: 98.9%; Sensitivity: 98.0%;
Zhen et al[17]	Retrospective	Hospital	2020	LC	China	31608 Images	CNN	AUC: 0.946
Giordano et al[53]	Retrospective	Hospital	2020	LC	Italy	167 Cases	SVM, RF	Accuracy: Exceeded 94%
Jeong et al[54]	Retrospective	Cancer Institute	2020	LC	China	1421 Cases	DL	AUC: 0.84
Appelbaum et al[31]	Retrospective	Hospital	2020	PDAC	America	594 Cases	LR	AUC: 0.71
Marya et al[44]	Retrospective	Hospital	2020	PDAC	America	1174461 EUS Images	EUS-CNN	Sensitivity: 99%; Specificity: 98%
Tonozuka et al[55]	Retrospective	Hospital	2020	PDAC	Japan	920 Cases	EUS-CAD	AUC: 0.94

GC: Gastric cancer; LC: Liver cancer; PDAC: Pancreatic ductal adenocarcinoma; DCNN: Deep convolutional neural network; CNN: Convolutional neural network; DL: Deep learning; SVM: Support vector machine; RF: Random forest; LR: Logistic regression; EUS-CNN: Endoscopic ultrasound convolutional neural network; EUS-CAD: Computer-assisted diagnosis system using deep learning analysis of EUS images; AUC: area under the curve; WLI: White-light imaging; NBI: Nonmagnifying narrow-band imaging.

Table 2 Recent researches on artificial intelligence in benign digestive system diseases

Ref.	Study design	Study population	Year	Disease	Country/Region	Number of cases	Methods	Results
Maeda et al[35]	Retrospective	Hospital	2019	UC	Japan	12900 Images	CAD	Sensitivity: 74%; Specificity: 97%; Accuracy: 91%
Popa et al[56]	Retrospective	Hospital	2020	UC	Romania	55 Cases	ML	Accuracy: 90%; AUC: 0.92
Tong et al[57]	Retrospective	Hospital	2020	UC	China	6399 Cases	RF, CNN	Sensitivity: RF (0.89); Sensitivity: CNN (0.90)
Qiu et al[38]	Retrospective	Hospital	2019	SAP	China	263 Cases	ANN	Sensitivity: 80.99%; Specificity: 89.44%
Chen et al[42]	Retrospective	Hospital	2019	SAP	China	389 Cases	LR	Accuracy: 87.1%
Namikawa et al[45]	Retrospective	Hospital	2020	Gastric ulcer	Japan	720 Images	A-CNN	Sensitivity: 93.3%; Specificity: 99.0%; PPV: 99.1%
Steinbuss et al[46]	Retrospective	Hospital	2020	Gastritis	Germany	1230 Images	CNN	Accuracy: 84%; Sensitivity: 100%; Specificity: 93%
Ozawa et al[47]	Retrospective	Hospital	2020	Colorectal polyps	Japan	16418 Images	CNN	Sensitivity: 92%; PPV: 86%

UC: Ulcerative colitis; SAP: Severe acute pancreatitis; CAD: Computer-aided diagnosis; ML: Machine learning; RF: Random forest; CNN: Convolutional neural network; ANN: Artificial neural networks; LR: Logistic regression; A-CNN: Advanced convolutional neural network; AUC: Area under the curve; PPV: Positive predictive value.