Greek scientists develop an Artificial Intelligence system to diagnose COVID-19 rapidly

The AI system for the rapid diagnosis of coronavirus by CT-scan is developed by the National Network of Technology and Research Infrastructures

The National Network of Technology and Research Infrastructures (EDYTE) implements the development of a fast and reliable diagnostic method for the detection of coronavirus and COVID-19, based on Artificial Intelligence / Deep Learning and the analysis of chest CT-scans.

CT scan of the lungs of a coronavirus patient
An attending physician looks at a CT scan of the lungs of a coronavirus patient in the covid-19 patient nursing department of the HMC Westeinde hospital in The Hague, the Netherlands, 04 April 2020. EPA/REMKO DE WAAL EDITORIAL USE ONLY

The method can improve the diagnosis’s effectiveness, especially in the early stages, and the evaluation of patients’ treatment. The development of such a system can be an essential aid in the molecular RT-PCR test. For the system’s implementation, imaging examinations are used, which are stored in the Health Data Center of EDYTE (HARMONI project), according to a memorandum of cooperation signed with the Ministry of Health.

A set of training data with anonymized CT-scans is created in the context of the project, where each CT-scan consists of a sequence of two-dimensional digital images. The data set is used to train a deep learning system, so that by “reading” an CT-scan, the system will diagnose whether or not there are signs of COVID-19. At the same time, the system discloses further information about the data that led to it’s diagnosis.

The current set of training data includes 1,500 CT-scans, which is continuously increasing, since the ability of such a deep learning system to provide the right diagnosis is based on its training with a large and representative sample data of COVID-19 cases in our country.