Russian AI predicts death risks and provides personalized analysis for diabetic patients
Archive photo / obltv.ru
This model can also explain its conclusions to specialist physicians, according to a statement from the university's public relations office.
The statement indicated that the model was trained on the body status indicators of over 550 diabetic patients monitored for 17 years. From hundreds of clinical and laboratory criteria, the artificial intelligence selected ten key biomarkers that constitute the long-term predictor. The key advantage of this development lies in the interpretation of neural network analysis using the SHAP method, which clarifies the critical data for prediction, achieving an accuracy rate of 84% in predicting survival throughout the monitoring period.
The study's author and director of the Institute of Aging Biology Research, Mikhail Ivanchenko, stated that the novelty of the research lies in the development of an accurate and interpretable predictive tool. He said, "Thanks to the method of interpreting AI predictions, it is possible to identify the interrelationships between dozens of patient status parameters. For example, it was found that the patient's age, disease duration, and number of complications are the strongest risk factors for death from diabetes. The approach also allows for the creation of a personalized risk map for each patient, such as showing that the 68% increased risk of death is mainly due to elevated creatinine, age, and four diabetic complications."
The development also points to the importance of some lesser-known biological markers, such as sodium brain peptide hormone (NT-proBNP) which reflects the hidden stress of the heart muscle, creatinine which indicates the condition of the kidneys, and a specific structure of N-glycan in blood serum as a biomarker of immune regulation and aging processes.
In this way, explainable artificial intelligence is transformed from an abstract algorithm into a practical tool that enhances the clinical thinking of the physician, an important step towards prolonging and improving the lives of millions of people living with diabetes.
