Explainable-by-design Semi-Supervised Representation Learning for COVID-19 Diagnosis from CT Imaging

• DOI: 10.48550/arxiv.2011.11719
• Publication Date: 2020-01-01
• Authors: Berenguer, Abel Diáz; Sahli, Hichem; Joukovsky, Boris; Kvasnytsia, Maryna; Dirks, Ine; Alioscha-Perez, M...; Deligiannis, Nikos; Gonidakis, Panagi...; Sánchez, Sebastiá...; Brahimetaj, Redona; Papavasileiou, Ev...; Chana, Jonathan C...; Li, Fei; Song, Shangzhen; Yang, Yixin; Tilborghs, Sofie; Willems, Siri; Eelbode, Tom; Bertels, Jeroen; Vandermeulen, Dirk; Maes, Frederik; Suetens, Paul; Fidon, Lucas; Vercauteren, Tom; Robben, David; Brys, Arne; Smeets, Dirk; Ilsen, Bart; Buls, Nico; Watté, Nina; Mey, Johan de; Snoeckx, Annemiek; Parizel, Paul M.; Guiot, Julien; Deprez, Louis; Meunier, Paul; Gryspeerdt, Stefaan; Smet, Kristof De; Jansen, Bart; Vandemeulebroucke...
• Keywords: FOS: Computer and information sciences; semi-supervised learning; diagnosis; Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV); Computer Science - Computer Vision and Pattern Recognition; COVID-19; imaging; 02 engineering and technology; Electrical Engineering and Systems Science - Image and Video Processing; 3. Good health; FOS: Electrical engineering, electronic engineering, information engineering; 0202 electrical engineering, electronic engineering, information engineering; eess.IV; cs.CV; CT

Our motivating application is a real-world problem: COVID-19 classification from CT imaging, for which we present a explainable Deep Learning approach based on a semi-supervised classification pipeline that employs variational autoencoders to extract efficient feature embedding. We have optimized the architecture of two different networks for CT images: (i) a novel conditional variational autoencoder (CVAE) with a specific architecture that integrates the class labels inside the encoder layers and uses side information with shared attention layers for the encoder, which make the most of the contextual clues for representation learning, and (ii) a downstream convolutional neural network for supervised classification using the encoder structure of the CVAE. With the explainable classification results, the proposed diagnosis system is very effective for COVID-19 classification. Based on the promising results obtained qualitatively and quantitatively, we envisage a wide deployment of our developed technique in large-scale clinical studies.