Unsupervised Class-Expert Learning for Supporting Covid-19 Triage Based on Computed Tomography Data

Deep learning applications in medical imaging have been achieving promising results the detection of diseases, among which clinical trials terms screening and diagnosis patients with COVID-19 stand out. Computed Tomography (CT) images chest used by specialists for COVID-19. However, due to need moment possibility using computational resources help team, it is observed literature several proposed works supervised learning, however lacks unsupervised methods In this work, deep models Convolutional Neural Network (CNN) Variational Autoencoders are feature extraction later information binary multiclass classification (k-means, Fuzzy C-Means Self-Organizing Maps). For purpose, a public database containing 4173 CT (2168 slices from COVID-19, 758 Healthy 1247 other lung diseases) was used. The show that via has similar performance state-of-the-art mainly accuracies 95.9%, 92.1% 95.9% k-means, SOM, respectively, presenting competitive literature. It also shows importance extracting features through convolutional networks improve performance, resulting use its state art area computer vision.