COVID-19 Detection Ensemble Analysis with Advanced Feature Descriptors (CODEX-AFD) Using Machine Learning Techniques

Infectious diseases like Covid-19 continue to pose a significant global health risk, with the potential to cause widespread pandemics. Precise prediction and detection of Covid-19 present formidable challenges in medical analysis, exacerbated by its capacity to induce a myriad of post-complications, including lung infections, cardiovascular issues, and post-traumatic disorders. Machine learning algorithms have emerged as invaluable tools in aiding the early detection of infectious diseases. In this research, a novel approach Covid-19 detection with advanced feature descriptor (CODEX-AFD) model is introduced for the detection of Covid-19. The proposed methodology incorporates Histogram Oriented Gradients (HOG) and Local Binary Pattern (LBP) as feature extraction techniques, combined with Random Forest (RF) classifier to classify CT scan images into Covid and Normal classes. This research extensively analyzed and compared the performance of LBP-RF and HOG-RF models with a proposed model. Evaluation metrics such as accuracy, precision, recall, and F1 score were utilized. Furthermore, the proposed model is compared with four different hybrid deep learning methods including EfficientNetB0-RF, EfficientNetB3-RF, VGG16-RF, and VGG19-RF models. The findings indicate that the suggested model outperforms both LBP and HOG-based Random Forest classifiers, exhibiting superior performance compared to hybrid models, especially in the detection of Covid-19. This suggests promising potential for assisting radiologists in automated diagnosis using CT scan image.