Efficient Deep Learning Classification of Cardiovascular Diseases Using Continuous Wavelet Transform and Improved Neural Network

An electrocardiogram (ECG) is an important diagnostic tool used to diagnose cardiovascular diseases (CVDs). Several computer aided diagnostic tools have been proposed using various machine learning algorithms, but they are facing challenges when the dataset is increased. This study proposes an efficient deep learning technique for classification of CVDs. Initially the method uses Continuous Wavelet Transform (CWT) to pre-process raw ECG signals. It embraces data equalization, denoising, segmentation and calculation of filter coefficients. In the second step, filter coefficients are converted into images using scalogram technique and these images are applied to an improved deep learning neural network. The neural network automatically computes and select the required features and hence uses them for classification. Lastly, the data is classified into three classes, such as Arrhythmia (ARR), Congestive Heart Failure (CHF) and Normal Sinus Rhythm (NSR) using Rectified Linear Unit (ReLu) activation function and an improved learning rate based VGG16 deep learning network. The recommended method can detect CVDs with a high accuracy using MATLAB. The suggested method is able to achieve admirable performance with improved learning rate, attaining an accuracy of $99.44 \%$, sensitivity of $99.45 \%$, precision of $99.44 \%$, and F-measure of $99.72 \%$. The proposed method was compared with AlexNet, hybrid AlexNet-SVM, CNN-LSTM, LDA and hybrid AlexNet-NCA-SVM and it was proved to be more accurate.