An efficient mixture of deep and machine learning models for COVID-19 diagnosis in chest X-ray images

• DOI: 10.1371/journal.pone.0242535
• Publication Date: 2020-11-17T18:30:09Z
• Authors: Dingding Wang; Jiaqing Mo; Gang Zhou; Liang Xu; Yajun Liu
• Keywords: Transfer of learning; Robustness

A newly emerged coronavirus (COVID-19) seriously threatens human life and health worldwide. In coping fighting against COVID-19, the most critical step is to effectively screen diagnose infected patients. Among them, chest X-ray imaging technology a valuable diagnosis method. The use of computer-aided images COVID-19 cases can provide experts with auxiliary suggestions, which reduce burden certain extent. this study, we first used conventional transfer learning methods, using five pre-trained deep models, Xception model showed relatively ideal effect, diagnostic accuracy reached 96.75%. order further improve accuracy, propose an efficient method that uses combination features machine classification. It implements end-to-end model. proposed was tested on two datasets performed exceptionally well both them. We evaluated 1102 images. experimental results show + SVM as high 99.33%. Compared baseline model, improved by 2.58%. sensitivity, specificity AUC 99.27%, 99.38% 99.32%, respectively. To illustrate robustness our method, also another dataset. Finally achieved good results. related research, has higher classification performance. Overall, substantially advances current radiology based methodology, it be very helpful tool for clinical practitioners radiologists aid them in follow-up cases.