Multi-Layered Architecture Convolution Neural Networks for Diagnosing and Predicting Heart Diseases on Multi-Modal

Abstract

Recently, heart diseases are increasing daily with the current lifestyle. Different types of cardiac diseases need to be diagnosed accurately for faster treatment to avoid death rates. Heart diseases are susceptible to attack a person speedily and cause sudden deaths. Medical experts are confused while diagnosing a person who comes in a medical emergency. In-time diagnosis, accurate prediction, and the proper treatment can save patients from heart diseases. Heart-related diseases cannot be predicted only by ECG data related to other medical information of a patient. This paper aims to develop an advanced artificial intelligence algorithm for analyzing and detecting heart diseases on multi-modal data. One of the advanced machine learning algorithms, Convolution Neural Network, is implemented with different architectural configurations for analyzing Body Mass Index (BMI), Electrocardiogram (ECG), and Physikalisch-Technische Bundesanstalt (PTB). Multiple CNN models are used for analyzing the multi-modal patient data to predict the presence of heart diseases. The CNN models are created by 80% of the training data and validated by 20% of the testing data. Different hyperparameters of the CNN models are tuned well to extract more data features to improve the prediction accuracy. The CNN models are developed in Python for processing and categorizing the feature variables as to whether heart disease is present or not. Medical data analytics require this model to get 98% of prediction accuracy.