Diabetes Detection Using ECG Signals: An Overview

Abstract

Diabetes Mellitus (or diabetes) is a clinical condition marked by hyperglycaemia and it affects a lot of people worldwide. Hyperglycaemia is the condition where high amount of glucose is present in the blood along with lack of insulin. The incidence of diabetes affected people is increasing every year. Diabetes cannot be cured. It can only be managed. If, not managed properly, it can lead to great complications which can be fatal. Therefore, timely diagnosis of diabetes is of great importance. In this chapter, we see the effect of diabetes on cardiac health and how heart rate variability (HRV) signals give an indication about the existence and acuteness of the diabetes by measuring the diabetes-induced cardiac impairments. Extracting useful information from the nonstationary and nonlinear HRV signal is extremely challenging. We review that deep learning methods do that extricating task very effectively so as to identify the correlation between the presence of diabetes and HRV signal variations in the most accurate and fast manner. We discuss several deep learning architectures which can be effectively used for HRV signal analysis for the purpose of detection of diabetes. It can be seen that deep learning methods is the state of art to understand and analyse the fine changes from the normal in the case of HRV signals. Deep learning networks can be developed to a scalable framework which can process large amount of data in a distributed manner. This can be followed by application of distributed deep learning algorithm for learning the patterns so as to do even correct predictions about future progress of the disease. Presently, there is no publicly available data of normal and diabetic HRV. If large amount of private data of diabetic HRV and normal HRV can be made available, then deep learning networks have the capability to give the authorities different kind of statistics from the stored data and projections of future prognosis of diabetes.