COVID-VIT: classification of Covid-19 from 3D CT chest images based on vision transformer model

This paper presents an explainable deep learning network to classify COVID from non-COVID based on 3D CT lung images. It applies a subset of the data for MIA-COV19 challenge through the development of 3D form of Vision Transformer deep learning architecture. The data comprise 1924 subjects with 851 being diagnosed with COVID, among them 1,552 being selected for training and 372 for testing. While most of the data volume are in axial view, there are a number of subjects’ data are in coronal or sagittal views with 1 or 2 slices are in axial view. Hence, while 3D data based classification is investigated, in this competition, 2D axial-view images remains the main focus. Two deep learning methods are studied, which are vision transformer (ViT) based on attention models and DenseNet that is built upon conventional convolutional neural network (CNN). Initial evaluation results indicates that ViT performs better than DenseNet with F1 scores being 0.81 and 0.72 respectively. (Codes are available at GitHub at https://github.com/xiaohong1/COVID-ViT). This paper illustrates that vision transformer performs the best in comparison to the other current state of the art approaches in classification of COVID from CT lung images