Explainable Deep Learning Model for COVID-19 Screening in Chest CT Images

El Habib Daho, M.; Khouani, A.; Lazouni, M. E. A.; Mahmoudi, S. A.

In this work, we proposed an Explainable model based on Deep Learning for fast COVID-19 screening in chest CT images. We first collected a database of 360 COVID and Non-COVID images at the Tlemcen hospital in Algeria. This database was merged with two other public datasets (the first one has been collected from several articles published on medRxiv, bioRxiv, NEJM, JAMA, Lancet, etc. The second one was obtained from infected cases in hospitals in Sao Paulo, Brazil). We also conducted a comparative study between Deep Learning classification models that are widely used in the state of the art such as VGG16, VGG19, Inception v3, ResNet50, and DenseNet121. We also proposed an interpretable architecture based on the ResNet50 model and the GradCam explanation algorithm. Experimentations showed promising results and prove that the introduced model can be very useful for the diagnosis and follow-up of patients with COVID-19. © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.