SCNN-VGG 16: SIAMESE CONVOLUTIONAL NEURAL NETWORK-VISUAL GEOMETRY GROUP 16 FOR CARDIOVASCULAR DISEASE DETECTION USING ECG SIGNAL

Cardiovascular disease has been one of the primary causes of death worldwide for the past few years in both developing and developed countries. Accurate diagnosis of cardiovascular disease and constant monitoring of patients can lower the mortality rate. However, detecting heart disease accurately is important, but for some reason, it is not always possible in all cases because it requires significant time, specialized knowledge and expert skill. In order to overcome these issues, a novel approach is developed, namely Siamese Convolutional Neural Network-Visual Geometry Group (SCNN-VGG 16), to detect cardiovascular disease using the Electrocardiography (ECG) signal to improve diagnostic accuracy. At first, the ECG signal is fed into the clinical feature extraction process to extract specific features. Next, feature selection is performed using Matusita Distance to recognize the most relevant features. Consequently, the input ECG signal and selected features are used to detect cardiovascular disease, which is performed by using the SCNN-VGG 16 model. In addition, the proposed SCNN-VGG 16 is an integration of Siamese Convolutional Neural Network (SCNN) and Visual Geometry Group (VGG 16), where Taylor concept is used for layer modification. The performance of the proposed SCNN-VGG 16 model is assessed using metrics, including accuracy, sensitivity and specificity, which acquired maximum values of 91.906%, 90.990% and 90.866%, respectively.