Medical Image Analysis with Transformer Models: A Study on Vision Transformers for the COVID19 Classification of CT Chest Images

International Symposium on Innovative Approaches in Smart Technologies (ISAS), pp. 1–5

Abstract

In this work, the application of transformer models, in particular a pre-trained vision transformer (ViT), for the classification of COVID-19 cases based on CT scans of lungs is investigated. Based on the COVIDx dataset, different training settings were evaluated to improve the performance of the model. The results show that the Vision Transformer was able to achieve a high accuracy of $\mathbf{0. 9 7 8}$ after only a few epochs, although the original dataset was heterogeneous.

Authors 2

1. Lena Charlotte Altherr Aachen Faculty of Electrical Engineering

   FH Aachen

   Affiliation as printed

   Aachen University of Applied Sciences,Faculty of Electrical Engineering,Aachen,Germany

2. Patrick Philipp

   Koblenz University of Applied Sciences

   Affiliation as printed

   Koblenz University of Applied Sciences,Faculty of Mathematics and Technology,Koblenz,Germany