OBJECTIVES: To apply advanced artificial intelligence to create an app for automated detection and prediction of gestational diabetes mellitus.METHODS: We applied five advanced artificial-intelligence algorithms (Neural network, SVM, Random forest, Adaboost, and MLP) to 3000 cases from a maternal-patient database with three demographics and one clinical outcome (gestational diabetes, GDM, which was diagnosed by ADA2011 with the oral glucose tolerance test). Clinical data are collected from one university-based regional hospital from South China. PCA and dimension reduction were used for data processing to identify the most critical parameters from the parameters. 2,500 of the 3,000 clinical cases were chosen as a training set and tested separately through Random Forest. The other 500 cases were tested with the model and used to develop an app for GDM screening. The sensitivity and specificity were calculated. Detection performances of this algorithm were evaluated by using the area under the receiver operating characteristic curve (AUROC).RESULTS: Detection performances of the five algorithms were a range of AUROC. The AUC for Neural network, SVM, Random Forest, Adaboost, and MLP was 0.7987, 0.7979, 0.7999, 0.7732 and 0.7974, respectively. For the Random Forest algorithm, the sensitivity for detecting GDM is 75%, which is better than 62% (without algorithm), using fasting glucose in a low medical resource condition.CONCLUSIONS: Random Forest algorithm had a high sensitivity for predicting gestational diabetes mellitus. This automatic detection algorithm outperformed doctors in prediction performance for gestational diabetes mellitus on fasting blood glucose in a low medical resource condition. The artificial intelligence application can promote long-distance medical care while possibly improving the quality and efficiency of maternal health care in rural areas. Artificial intelligence medical applications have a promising future, and more research on efficacy, safety, and cost-effectiveness should be conducted.