Hybrid FNN-DNN Framework for Early Detection of Cardiac Arrhythmia: A Comprehensive Approach for Enhanced Diagnostic Accuracy

Abstract

Cardiac arrhythmias constitute a major cause of morbidity and mortality in most settings of the world, and early identification of arrhythmias is paramount in efficient treatment and enhanced patient care. The widely used traditional electrocardiogram (ECG) interpretation, though clinically useful, is also subject to dysfunctionality in the form of its subjectivity of interpretation and lack of sensitivity to finer entrances of abnormalities. In the proposed study, a hybrid of a Feedforward Neural Network and Deep Neural Network (FNN and DNN) architecture system is proposed to significantly improve the early diagnosis of cardiac arrhythmia with structured clinical data contained in a publicly accessible Heart Disease Dataset. The suggested scheme leverages the effectiveness of FNN in handling structured attributes, combined with the ability of DNN to model multilateral and complex intricacies in characteristics, thereby facilitating the complete representation of the characteristics. The hyperparameter tuning and regularization were performed on the structure of the hybrid architecture, resulting in optimization, and accuracy, precision, recall, F1-score, and AUC-ROC leading metrics were adopted. Experimental outcomes reflected that the accuracy of 84.8% was as good as the standalone FNN and DNN models, displaying a balance in the performance of all the considered metrics. Analysis of the confusion matrix has shown a high level of classification reliability with no notable bias over one of the classes. The ultimate contribution that can be made using the study is a computationally efficient and generalizable hybrid model, which can be incorporated into the clinical workflow and electronic health records (EHR) systems. The modular nature of its design contributes to future extension, such as the analysis of the raw ECG signals and explainable AI. The results show that the hybrid FNNDNN holds promise as a scalable, easily interpretable, and accurate device to proactively detect arrhythmia, allowing for more accurate care and cardiovascular diagnostic outcomes.