Gastric cancer susceptibility mapping using GIS and machine learning in East Azerbaijan province, Iran.

Despite technological advancements, Gastric Cancer (GC) remains one of the most lethal malignancies worldwide. This study aimed to map spatial risk susceptibility for GC in East Azerbaijan province, located in northwest Iran, which ranks second nationally in GC incidence. In the first step, the residential addresses of 2,884 patients diagnosed with GC between 2015 and 2019 were geocoded using Google Earth, converting them into geographic coordinates for spatial analysis. The dataset was subsequently divided into two subsets: 80% of the cases were allocated for model training and 20% for validation. A set of environmental variables including elevation, rainfall, iron, volcanoes, copper, coal, Pb-Zn, arsenic, and mercury along with two models of Logistic Regression (LR) and Logistic Model Tree (LMT) was used to map areas susceptible to GC. The performance of the models was evaluated using Area Under the Curve (AUC) and Root Mean Square Error (RMSE) methods. The findings revealed that central regions of the study area are the hotspots for occurrence of the disease. Both LR and LMT models pinpointed nine counties as high-risk and susceptible areas for GC, while five counties were categorized as low-risk areas. In addition, evaluation of RMSE and AUC metrics showed that the LR model outperformed the LMT model across the study area. More results revealed that spatial clustering of GC cases in areas with higher rainfall and closer proximity to heavy metal deposits, supporting the hypothesis that environmental exposure contributes to disease risk. This study is useful for environmental health researchers, and regional public‑health authorities to locate high‑risk areas, prioritize screening, and allocate diagnostic and treatment resources efficiently.