CovidStopHospital: e-Health Service for X-Ray-Based COVID-19 Classification and Radiologist-Assisted Dataset Creation

Abstract

Image data processing using artificial intelligence (AI) algorithms has various applications, including medicine. During the SARS-CoV-2 pandemic, many successful COVID-19 classification algorithms were trained. However, to be effectively used in clinical settings, these algorithms need to be deployed in hospitals. Existing platforms for AI algorithm deployment may not be usable in hospitals that rely on proprietary information systems lacking application interfaces. This paper introduces an easily modifiable general AI-X-ray service capable of deploying AI algorithms even in hospitals using proprietary information systems lacking application interfaces. The CovidStopHospital service, based on the AI-X-ray architecture, is also presented. It is designed for COVID-19 classification and can seamlessly incorporate any classification AI algorithm; the presented solution uses DeepCovid-XR algorithm. The service also includes functionality for radiologists to label X-ray images, facilitating the creation of new datasets. CovidStopHospital underwent testing to ensure its stability and performance, with an average X-ray analysis time of 11.53 seconds and a maximum of 14.01 seconds. The tool can potentially be a valuable diagnostic support tool and is currently in experimental deployment at the University Hospital of Olomouc.

Ähnliche Arbeiten