Role of artificial intelligence in screening and medical imaging of precancerous gastric diseases

Table 1 List of studied artificial intelligence models

Number	Models	Area of application
Laboratory models
1	Machine learning and deep learning models (logistic regression, XGBoost, LightGBM, random forest, support vector machine and CatBoost)	Serological control arbovirus infections such as Dengue, Chikungunya and Zika in Latin America, South America and Pakistan
2	Automated machine learning platform MILO	Monitoring active tuberculosis
3	Machine learning model (XGBoost + AdaBoost)	Early detection of esophageal adenocarcinoma and Barrett's esophagus
4	Machine learning models (BW and BW + MAT)	Serological screening for leptospirosis
5	Three machine learning models, random forest, support vector machine and AdaBoost	Prediction of hepatitis C infection
Endoscopic models
6	Deep convolutional neural network system ENDOANGEL	Endoscopic diagnostics of atrophic gastritis, intestinal metaplasia
7	AIMED model that uses MobileNetV3-large convolutional neural networks	Identification of the area of gastric anatomy
8	EGC-YOLO artificial intelligence model	Detecting gastric cancer at the early stage of cancer
9	MMP-AI system	Detection of submucosal tumors, gastric leiomyomas, gastric stromal tumors and ectopic gastric pancreas
10	GAM-Efficient Net Neural Network	Diagnosis of atrophic gastritis and intestinal metaplasia
Histological models
11	Artificial Intelligence Model Aiforia Technologies, Helsinki, Finland	Diagnosis of eosinophilic esophagitis
12	Deep learning neural network Attach-ANN (trained using WSI BCC)	Diagnosis of basal cell carcinoma of the skin
13	ImageNet convolutional neural networks (Inception ResNet v2, Inception v3, ResNet152, Xception) based on Vision Transformer deep machine learning model	Salivary gland neoplasm diagnostics
14	Graph neural network SpaInGNN	Transcriptomics technology that allows characterizing gene expression patterns in tissue
15	GasHisSDB	Gastric histopathology

XGBoost: Extreme gradient boosting; MILO: Machine Intelligence Learning Optimizer; BW: Blood work models; MAT: Model, BW + MAT, incorporated the initial hospitalization MAT titer in addition to the patient features and clinicopathologic parameters; MMP-AI: Multimodal multipathway artificial intelligence.