Deep Learning-Powered IoT Wearables for Early Detection of Cardiovascular Diseases

Abstract

Cardiovascular illnesses (CVDs) are still the top cause of death around the world. To improve prognoses and lower healthcare costs, early monitoring systems are needed. This study introduces a new system that combines deep learning algorithms with smart IoT devices to make it easier to track and find CVDs early in real time. Wearable monitors collect constant physiological data like heart rate, blood pressure, oxygen levels, and electrocardiogram (ECG) readings that are used in the suggested system. We used a CNN-LSTM design that combines a Convolutional Neural Network and Long Short-Term Memory to handle the signals' time and spatial patterns. The CNN part pulls out important features from raw, multidimensional sensor data, and the LSTM part finds time relationships to make predictions more accurate. The dataset used includes PhysioNet's publicly available cardiovascular health records and real-time data from smart devices. To balance class distributions, simulated minority oversampling was added to the dataset. Precision, memory, F1-score, and accuracy were used as measures to evaluate performance. It did better than standard models like CNN, LSTM, and classic machine learning classifiers, with a total accuracy of 96.4%, a precision of 95.7%, and an F1-score of 96.1% in finding early signs of CVD. The system also allows implementation at the edge, which ensures low delay and energy-efficient processing that is good for constant tracking. The results show that combining deep learning with IoT gadgets could greatly improve early identification of CVD, allowing for proactive actions and personalised healthcare. The suggested structure offers an adaptable and affordable way to keep an eye on people's heart health in real time, especially in places that are far away and don't have a lot of resources.