Deep Learning–Enabled Cardiovascular Disease Detection using DenseNet

Cardiovascular diseases pose a global health threat today. Diagnosis needs to be performed quickly so; therefore it is important for healthcare providers to use reliable methods of diagnosis. ECG signals play an important role in measuring the health of the heart. Interpreting the signal manually takes time and is also subject to inconsistent results by varying medical practitioners. Machine learning approaches that have been developed for ECG signal interpretation do not adequately capture the intricate and subtle variations in the ECG's waveform. The goal of the research is to develop an intelligent deep learning framework that use ECG image data to estimate cardiovascular risk. A densely connected convolutional neural network, DenseNet was utilised and is capable of extracting many detailed morphological features from ECG images because of its capability of reusing its features efficiently and propagating gradients effectively. When ECG images are prepared for analysis, the images undergo various pre-processing operations to normalise the input, reduce noise and augment the data set to improve consistency and robustness. Using the trained model, many cardiac conditions such as arrhythmias, coronary artery disease, and heart failure can be found based on the visual pattern of the ECG. Dense connectivity helps to better classification dependability and reduce the possibility of overfitting when using a wide variety of datasets.