MACHINE LEARNING MODELS FOR ANALYZING DATA FROM GLUCOMETERS AND HEMATOLOGY ANALYZERS IN MEDICAL DIAGNOSTICS TASKS

Abstract

With the global increase in the incidence of diabetes and anemia, there is an increasing need to develop modern, accurate and scalable tools for early diagnosis and continuous monitoring of these conditions. The purpose of this study is to develop an approach to integrating data obtained from modern express glucose meters, hematological and biochemical analyzers with machine learning algorithms for automated classification and assessment of the risk of diabetes and anemia. The article uses data collected from medical devices, including glucose, glycated hemoglobin (HbA1c), ferritin, hematocrit, MCV, MCH, MCHC and other biomarkers. The data was standardized, processed, and used for training and testing machine learning models: DecisionTree, randomForest, AdaBoost, ExtraTrees, GradientBoosting, as well as logistic regression, SVM, and XGBoost. Additionally, model interpretation methods such as feature importance and SHAP were used. The results showed that the best metrics (accuracy 0.92, F1-metric 0.91) were achieved using the GradientBoosting model with hyperparameter presetting. The model also showed high stability and interpretability, which is critical for use in clinical practice. The implementation of such a system can significantly improve the efficiency of diagnostics, reduce the burden on medical staff and provide a personalized approach to patient monitoring.  The study confirms the high prospects of using AI in laboratory diagnostic tasks and highlights the need for further integration of medical devices with intelligent analytical platforms.