Accurate Approach to Diabetes Detection Using Deep Learning Algorithms

Abstract

Diabetes is a widespread metabolic disorder impacting millions globally. Its occurrence increasing at an alarming rate each year. Diabetes needs to be properly managed it can lead to severe and potentially fatal complications in various vital organs. Early detection is crucial to initiating timely treatment, which can prevent the progression of the disease to such severe complications. The HRV signals can therefore be used as a non-invasive approach for identifying diabetes. Variations in the time intervals between the heartbeats provide crucial information regarding the efficiency of the autonomic nervous system, which varies significantly in diabetic patients.  This paper introduces a novel approach for classifying diabetic and non-diabetic HRV signals using advanced deep-learning techniques. We use LSTM networks, CNNs, and combined models to extract detailed temporal features of HRV. The extracted features are then used and fed into a support vector machine (SVM) for the classification process, which aims at differentiating between diabetic and non-diabetic signals of HRV.