Detection of COVID-19 by classifying CT-Scan images using Enhanced MobilenetV2
Main Article Content
Abstract
Using images from chest CT scans, this study attempts to determine how well transfer learning performs in correctly identifying COVID-19. This research enhanced the MobileNetV2 deep learning model, decision trees, support vector machines, and logistic regression for feature extraction and classification, respectively. 13545 CT scan images of the chest were included in the dataset for this study. A smaller dataset of images from pneumonia, COVID-19, normal, Omicron and delta chest CT scans is used to fine-tune the models after they have been trained on a larger image dataset. When machine learning models based on transfer learning are applied to chest CT scan images, the study's findings demonstrate that COVID-19 can be successfully identified. The MobileNetV2 model turned out to be the top performer. This model performed well on the testing set, achieving a 97.5% accuracy rate along with good precision, recall, and F1-score across all classes. Overall, the model performed well, as evidenced by the high average precision and recall. As a result, the study shows how well deep learning works for both image classification and feature space extraction from chest CT scan images used for COVID-19 detection.