Multimodal machine learning approaches for heart disease detection and control

Abstract

Heart disease is one of the major causes of death around the world. This is due to the high and rising number of smokers and other factors. Developing a novel system capable of predicting heart diseases is necessary to control this health hazard. An electrocardiogram (ECG) dataset is utilized to build the machine learning-based model. The ECG is widely used in the clinical diagnosis of cardiac conditions. It is a dependable metric for tracking the cardiovascular system’s performance. Because ECG signals are easy to utilize and noninvasive, they have been employed extensively for the detection of heart problems. Four feature extraction methods were integrated to extract important features from the ECG signals. This research can be extended to a multimodal architecture combining ECG with complementary physiological signals (e.g., PPG, SpO₂, HRV, and clinical biomarkers) to increase diagnostic reliability. Secure deployment can be achieved using federated learning and blockchain-based authorization, while low-power inference enables sustainable integration into wearable and telehealth environments. The extracted features were used to train and test—support vector machine classifier model. The proposed system is proven to be very accurate and cost-effective. This paper presents a machine learning-based system to predict heart abnormalities using features extracted from ECG samples. Predetermination and timely detection of heart abnormalities by the proposed system are important to save as many lives as possible.