Review of Neural Networks for Prediction and Diagnosis of Cardiovascular Diseases

While the continuous recording of heart signals allows for timely monitoring, prediction, and alerting, heart diseases are among the leading causes of global mortality. Deep learning algorithms, capable of processing diverse inputs such as images, audio, and text, are utilized in various medical fields including disease diagnosis, health monitoring, and mortality prediction. The CNN-۱D network achieved an accuracy of ۹۹.۲% and a sensitivity of ۹۹.۶۲% on the UCI dataset by extracting hierarchical features from time signals, while the MLP model reached an accuracy of ۹۷.۸% but had a low sensitivity of ۵۶.۴۵% in identifying true positive cases. The CNN+PSO method offered an accuracy of ۹۲.۳۴% with lower complexity, aiming to reduce dimensions. The CNN+Bi-LSTM model also provided a desirable accuracy in estimating heart rate with MAE=۰.۹۰. This review article evaluates the performance of models based on accuracy, sensitivity, and specificity indicators in a comparative table format and provides an analysis of the proposed methods.

کلیدواژه ها:

Cardiovascular Disease Diagnosis ، Early Disease Detection ، Artificial Neural Network ، Multilayer Perceptron (MLP) ، Convolutional Neural Network (CNN-۱D) ، Long Short-Term Memory (LSTM) ، Electrocardiogram Signal (ECG)