Deep Learning-Based Hybrid CNN Model for Heart Disease Diagnosis Using ECG Images
Abstract
As is known, data is the most important part of healthcare, so data also plays an important role in working with heart diseases. In this study, a deep learning-based approach was developed and analyzed to classify features extracted from ECG images. In this study, ECG images were pre-processed at the initial stage, and the image was converted from color to grayscale, which reduces unnecessary features in the image and helps to train the model faster and more accurately. The dataset used in our study is divided into four main categories, each of which represents different states of the heart. Three models simple CNN, AlexNet, and SqueezeNet architectures were tested during the “train” and “test” processes on these 4-class datasets, and a hybrid model is proposed by combining the AlexNet and SqueezeNet models in order to achieve high accuracy and computational efficiency. The results show that the proposed hybrid model is 1.5% more accurate than the existing AlexNet model and 1.6% more precise than the existing AlexNet model. This hybrid model is also 0.4% more accurate than the SqueezeNet model and 0.5% more precise than the existing SqueezeNet model.