Abstract

Remote patient monitoring has become pivotal in managing chronic diseases like diabetes. This study proposes a novel approach for the classification of diabetes subtypes utilizing a deep-learning reconstruction algorithm. The system leverages continuous patient data obtained through remote monitoring devices, enabling real-time analysis for timely intervention. The deep-learning reconstruction algorithm, based on a convolutional neural network architecture, demonstrated exceptional accuracy in distinguishing between diabetes subtypes. The model achieved an overall classification accuracy of 92%, outperforming traditional methods. It exhibited high sensitivity and specificity, with values exceeding 90% for each subtype. The results showcase the system’s effectiveness in classifying diabetes subtypes: Type 1 diabetes (Sensitivity: 94%, Specificity: 92%), Type 2 diabetes (Sensitivity: 91%, Specificity: 94%), and Gestational diabetes (Sensitivity: 93%, Specificity: 91%). The system’s ability to accurately identify these subtypes ensures personalized and targeted care for patients.