| U-Net- based VGG Classification for COVID Detection in Chest X-ray Images |
| Paper ID : 1025-ICEEM2025 (R2) |
| Authors |
|
Eman elsaid alaa *1, Salah Khamis2, Amira s. Ashour3 1Department of Electronics and Electrical Communications Engineering ,Faulty of Engineering, Tanta University, Tanta, Egypt 2Department of Electronics and Electrical Communications Engineering, Faculty of Engineering Tanta University Tanta, Egypt 3Department of Electronics and Electrical Communications Engineering, Faulty of Engineering Tanta University Tanta, Egypt |
| Abstract |
| Chest infection classification becomes essential since the appearance of COVID. Many artificial intelligence based systems were developed for primary diagnosis of COVID-19 and comprehensive disease management. In this study, a segmentation-classification computer-aided diagnostic system is proposed. A U-net segmentation stage followed by Vgg16 as a feature extractor from chest X-ray images is applied. A comparative study with other deep learning networks, including Vgg16, Densenet201, Mobilenetv2, Resent50 and Google-net, were conducted to prove the effectiveness of the proposed system. The feature selection using Minimum Redundancy Maximum Relevance (MRMR) technique is also employed to choose the significant feature. Then, the support vector machine (SVM) is applied for binary classification of images to normal, or covid19 cases. The experimental results proved the superiority of utilizing U-net with achieved performance 97.20%, and 98.97% Jaccard and dice, respectively. In addition, Vgg16 followed linear SVM achieved accuracy, sensitivity, specificity, F-score, and precision of 99.7%, 0.963, 0.993, 0.997, and 0.994, respectively |
| Keywords |
| Chest X-ray, U-net, semantic segmentation, deep learning, Vgg16, SVM, feature selection, MRMR. |
| Status: Accepted |