Advanced AI Techniques for Real-Time Blood GlucosePrediction in Diabetics: A Study Using Deep Learningand Genetic Algorithms

Diabetes mellitus is one of the most common chronic diseasesworldwide, with its prevalence expected to rise sharply. Projectionssuggest that by 2050, more than 1.3 billion people globally will beliving with diabetes, a significant increase from the current estimateof 529 million. A case of Type I diabetes is characterized by thepancreas failing to produce adequate amounts of insulin, which leadsto uncontrolled blood glucose levels. Traditionally, managementinvolves patient-administered insulin and monitoring blood glucoselevels (BGLs) based on dietary intake reported by the patient. This study introduces an innovative method that leverages advancedArtificial Intelligence (AI) techniques to continuously predict bloodglucose levels for the short term (+30 minutes) from the currentsituation. The techniques applied include Deep Learning with ArtificialNeural Networks (ANNs), Genetic Algorithms (GAs), and ReinforcementLearning. These methods analyzed both raw BGL data and additionalinformation derived from a Diabetic Dynamic Model of BGLs. The study’s preliminary evaluation used data from four virtual patientsgenerated by an open-source diabetes simulation tool and threereal diabetic patients using the DexCom SEVEN system. The resultsindicated that the knowledge-based approach significantly enhancedprediction accuracy, with Genetic Algorithms outperforming ANNs.Additionally, the integration of online learning and ReinforcementLearning, which adapt to emerging data patterns, further improvedpredictive accuracy. This advanced methodology demonstrates considerable potentialfor enhancing diabetes management by providing timely and preciseBGL predictions without direct patient input. Future studies involvinglarger cohorts of both Type I and Type II diabetic patients are necessaryto validate these promising results.