Table 2

Summary of the different deep learning methods by applications in gastric problems

Based methodTaskApplicationReference
Convolutional neural networkClassificationBenign or malign of stomach biopsy specimens79
Benign or malignant images58
Gastric cancer or non-cancer32
Gastritis or non-gastritis34
Helicobacter pylori-positive or helicobacter pylori-negative38
Helicobacter pylori–related gastritis, reactive gastropathy and histologically normal gastric mucosa40
Neoplasm or non-neoplasm80
Normal gastric images or early gastric cancer images60
Normal or abnormal gastric slow wave61
With and without histology-proven atrophic gastritis35
Convolutional neural network and residual neural networkBenign ulcer and gastric cancer42
Early gastric cancer,
advanced gastric cancer, high grade dysplasia,
low grade dysplasia or non-neoplasm
Convolutional neural network with
deep generalised multi-instance learning
Differentiation degree (poorly and well/moderately) and
lauren type (intestinal, diffuse and mixed)
Convolutional neural network and deep reinforcement learningGastric sites62
Residual neural networkGastric cancer type (intestinal type or diffuse type)52
Microsatellite instable or microsatellite stability53
Recurrent neural networkLive or dead probability51
Convolutional neural networkClassification/detectionBenign or malignant gastric ulcer/gastric ulcer43
HER2+ tumour, HER2 tumour or non-tumour/necrosis detection5
DetectionGastric cancer65
Gastritis or non-gastritis36
Lymphocyte or non-lymphocyte66
Normal mucosa, non-cancerous pathology, cancer63
Signet ring cell cancer29
Signet-ring cell carcinoma component intramucosal or advanced82
Gastric ulcer43
Necrosis detection5
Generative adversarial networkGenerationGastritis image generation56
Residual neural networkSegmentationGastric cancer54
Fully convolutional networkGastric tumour48
Recognise small cancerous tissues49