COVID-19 diagnosis system applied to chest X-ray images

A team of researchers from INESC TEC’s Centre for Biomedical Engineering Research (C-BER) developed a computer-assisted diagnosis system, in partnership with radiologists from the Vila Nova de Gaia/Espinho hospital centre (CHVNGE) and the Northern Region Health Administration (ARS Norte).

This system can help define the diagnosis and treatment strategies for each patient, by proposing a second opinion to radiologist or other non-specialised staff.

Six researchers at C-BER were in charge of developing the algorithm, supported by two radiologists in terms of radiological observation and follow-up.

The algorithm developed by INESC TEC is based on deep learning methods. The system automatically learns the most relevant elements of the image for diagnosis. More specifically, it analyses a significant amount of images representing the different signs of COVID-19, but also images of healthy patients or individuals with other conditions. With enough data, the system can learn the characteristics representative of a condition, thus enabling the automatic diagnosis.

There are already several studies on the use of support systems for medical diagnosis in these situations, but their clinical applicability is yet to be tested.

The CXR_AI4COVID-19 project lasted for five months and was supported by the RESEARCH4COVID-19 financing line (€29K), promoted by the Foundation for Science and Technology (FCT).