Post-COVID effect on heart after recovery based on hybrid EfficientNet-DBN with multilevel classification using ECG images

• Published in: EngMedicine 2024-09, Vol.1 (2), p.100021, Article 100021
• Main Authors: Basith Ali Khan, Mohammed Abdul, Reddy, Edara Sreenivasa
• Format: Article
• Language: English
• Subjects: Coronavirus; Deep belief network (DBN); Deep learning (DL); Heart disease; Restricted Boltzmann machine (RBM)
• ISSN: 2950-4899; 2950-4899
• DOI: 10.1016/j.engmed.2024.100021

The highly contagious and dangerous virus known as COVID-19 is caused by severe acute respiratory syndrome (SARS), which is disseminated worldwide. Heart disease is a major cause of death worldwide. One year after the COVID-19 pandemic, the risk of heart issues is considerable. Such heart issues involve irregular heartbeats, heart failure, which is the inability of the heart to pump correctly, and coronary disease, which builds up in the arteries and causes restriction in blood flow, heart attacks, etc. Therefore, classifying this disease at an early stage is crucial. Hence, post-COVID effect on the heart after recovery with multilevel classification was performed using the hybrid EfficientNet ​+ ​Deep Belief Network (EfficientNet ​+ ​DBN) designed in this study. The input image obtained from the dataset was delivered for binary image conversion and then allowed for feature extraction. Subsequently, first-level disease classification was performed using the hybrid EfficientNet ​+ ​DBN to detect whether the disease was in normal or abnormal conditions. If it is categorized as abnormal, a second-level classification is performed using EfficientNet-DBN, which classifies myocardial infarction and COVID-19 patients. The Pearson's correlation coefficient was used for the post-COVID correlation study. The experimental outcome showed that EfficientNet-DBN attained a maximum accuracy of 89.80%, sensitivity of 94.70%, and specificity of 91.80%.