Computer aided diagnosis of heart disease Through classification of ECG signals Using computational intelligence

Abstract

The categorization of arrhythmia type in ECG signals has been newlinea significant computational methodological problem for the last ten years. newlineTypical methods include the categorization and extraction of features. newlineThis study attempts to categorize the many types of arrhythmia that may newlinebe used to evaluate patients with cardiovascular disorders and to examine newlinecomputational techniques like pre-trained neural networks. The MIT-BIH newlineArrhythmia database is used to test the proposed approach together with newlineother publicly accessible resources. The time-domain and frequencydomain newlineof the ECG data signals are the retrieved characteristics in this newlinecase. A recording of the heart s activity called an ECG, which is a record newlineof the heart s pumping motion, is the recommended method for spotting newlinethese aberrant occurrences. However, since the ECG carries so much newlineinformation, it is quite challenging to extract the relevant information newlinethrough picture analysis. It is crucial to develop a system that is efficient newlinefor processing the vast amounts of data from ECG. Some image newlineprocessing methods are used to transform the ECG signal into an image. newlineThis study provides a hybrid deep learning-based method to enhance the newlineidentification and classification process. The DENSENET, VGGNET and newlineRESNET for the identification and categorization of Arrhythmia types newlinewere explored in this research. This work makes two contributions. The newlinefirst step in creating 2D images from 1D ECG data is automating noise newlinereduction and feature extraction. The CNN-LSTM model, which newlinecombines the CNN and LSTM models, is given on the basis of newlineexperimental data. To evaluate the effectiveness of the proposed CNNLSTM newlineapproach, we carried out an extensive study utilizing the widely newlineknown MIT BIH arrhythmia dataset. The findings show that the accuracy newlinex newlinerate of the suggested strategy is 99.10%. The suggested model also newlineexhibits average sensitivity and specificity of 98.35% and 98.38% newlinerespectively. These results are far better than those obtained via other newlinemethods and they will drastically