Akademik Veri Yönetim Sistemi
7th International İzmir Congress on Medicine, Nursing, Midwifery, and Health Sciences, İzmir, Türkiye, 21 - 23 Ocak 2025, cilt.9789952854558, ss.647-652
EFFECT OF ARTIFICIAL INTELLIGENCE APPLICATIONS ON DETERMINING THE RISK OF GESTATIONAL DIABETES MELLITUS
ABSTRACT
Gestational diabetes mellitus (GDM) is defined as “varying degrees of glucose intolerance is detected during pregnancy”. GDM causes many adverse maternal-fetal effects, such as hyperglycemia, hypoglycemia, hypertension, pre-eclampsia, polyhydramnios, macrosomia and cesarean delivery. Therefore, it is important to predict GDM in pregnancy. In recent years, the use of artificial intelligence (AI) that can predict the risk of GDM has increased, so that GDM can be predicted even in the first trimester of pregnancy. AI are reported to the best prediction performance compared to traditional statistical methods and oral glucose tolerance test results. The aim of AI in predicting GDM risks is to create a generalizable GDM risk predictor that can provide heterogeneity across different populations, rather than just risk groups. For this aim, machine learning (ML) models such as multivariate logistic regression (LR), random forest (RF), extreme gradient boosting (XGBoost) and light gradient boosting machine (LightGBM) were found to be the most commonly used AI models in studies. These AI models have a work mechanism that addresses demographic variables, biochemical indicators, medical and clinical history of pregnant women. Predictors of GDM identified using AI models were quality of life in early pregnancy, diastolic blood pressure, age, ethnicity, history of GDM, body mass index (BMI), family history of diabetes mellitus and number of pregnancies. Serum insulin and allantoic acid are potential biomarkers for early GDM detection.The most important predictors for the diagnosis of GDM were pre-pregnancy BMI and maternal abdominal diameter at the beginning of pregnancy and, fasting blood glucose and HbA1c at the end of the first trimester. In conclusion, AI models show high performance and will be a more selective and cost-effective screening method for GDM. Widespread use of artificial intelligence models can help in earlier prediction of GDM, timely treatment and prevention of adverse metarnatal-fetal effects.
Keywords: Gestational diabetes mellitus, artificial intelligence, machine learning.