Deep-Learning Supported Detection of COVID-19 in Lung CT Slices with Concatenated Deep Features
Document Type
Article
Publication Date
1-1-2023
Abstract
This research proposes and implements an automatic diagnostic scheme for detecting COVID-19 infection using lung CT slices to decrease the diagnostic burden. The proposed framework consists of (i) Image collection and preprocessing, (ii) Deep feature mining using the chosen scheme, (iii) Feature reduction and serial integration, and (iv) Classification and validation. A pre-trained deep-learning scheme is implemented in this scheme to obtain the necessary deep features from the CT slices selected and then to reduce these features by 50%. A CT image classification task is initially performed with SoftMax, and the outcome is then verified with other binary classifiers. Finally, we present and discuss the results of the proposed classification work using (i) single PDS and (ii) dual-deep features. With a single PDS, the Random Forest (RF) classifier provided a detection accuracy of 94%, and the K-Nearest Neighbor (KNN) classifier provided an accuracy of 99%.
Publication Title
Lecture Notes on Data Engineering and Communications Technologies
First Page Number
359
Last Page Number
369
DOI
10.1007/978-981-99-3432-4_28
Recommended Citation
Sivakumar, R.; Kadry, Seifedine; Krishnamoorthy, Sujatha; Balaji, Gangadharam; Nethrra, S. U.; Varsha, J.; and Rajinikanth, Venkatesan, "Deep-Learning Supported Detection of COVID-19 in Lung CT Slices with Concatenated Deep Features" (2023). Kean Publications. 312.
https://digitalcommons.kean.edu/keanpublications/312