Denis Eka Cahyani, Lucky Tri Oktoviana, Anjar Dwi Hariadi, Faisal Farris Setyawan, Samsul Setumin
Recently, the transmission of Coronavirus disease has not disappeared in the society. Rapid screening with high accuracy is needed to detect COVID-19 so that the virus does not spread more widely. Chest X-Ray (CXR) images may be utilized to detect COVID-19 infections. This research examines InceptionV3-BiLSTM to InceptionV3, Xception, and Xception-BiLSTM models for detecting COVID-19 using CXR. This research utilizes the Database of COVID-19 Radiographic Images, which includes three classes: COVID-19, Viral Pneumonia, and Normal. The data scenarios are divided into two types, namely scenario 1 containing original data and scenario 2 containing balanced data. The InceptionV3-BiLSTM model has the highest accuracy value in scenario 1 and scenario 2 data with accuracy values of 98.25% and 97.77%, respectively. Then the InceptionV3 model obtained the second-best accuracy value. Followed by the Xception-BiLSTM model and finally the Xception model in each data scenario. In all of the data scenarios, the InceptionV3-BiLSTM model has relatively higher precision, recall, and F1-measure values than the two competing models. So, the conclusion of this investigation is the InceptionV3-BiLSTM model can produce excellent outcomes utilizing CXR for COVID-19 detection. © 2024 Author(s).
Department of Mathematics, Faculty of Mathematics and Natural Sciences, Universitas Negeri Malang, Malang, Indonesia; School of Electrical Engineering, Universiti Teknologi MARA, Cawangan Pulau Pinang, Malaysia