Early Detection of Kidney Disease Using ECG Signals Through Machine Learning Based Modelling

Abstract

This paper introduces the idea of detecting the presence of kidney disease through machine learning based classification modelling, by processing the patient’s ECG signal. Recent studies and ongoing researches have showed that patients undergoing kidney problems start developing cardiac problems- scientifically known as the Cardio Renal Syndrome (CRS) which can lead to a sudden cardiac arrest in the last stages of their disease. Since cardio-vascular diseases and the chronic kidney disease is inter-related, this model can be used for patients undergoing cardio-vascular problems to determine whether their kidneys have been effected or not. If the Chronic Kidney Disease (CKD) can be diagnosed at an earlier stage, it may give the patient some time to help reverse the disease or at least slow its progression by taking necessary medical steps. For this model, digitized ECG data was collected from open access databases such as PTB (for kidney patients) and Fantasia (for healthy people) from Physionet Database (www.physionet.org) and the model was later validated using different data from the same online database. The validation process gave satisfactory results, as the model could successfully classify the users from being healthy or a kidney patient. In our study, we found an accuracy level of 97.6% which was the highest using both features QT and RR interval, in comparison to the accuracy that was found when either one of the features was used.