The goal is to create AI models to assist in the analysis of thoracic radiographies in patients in a context of emergency.

A consortium integrating Hospital da Luz Learning Health, Instituto Superior Técnico (IST) and INESC-ID has an innovative project in progress, with the purpose of developing a solution composed of artificial intelligence (AI) models for the analysis of thoracic radiographies undertaken by patients in hospital emergency units, estimating the probability of presenting the characteristics of covid-19 infection. In case of high probability, the models enable to estimate the degree of severity of the disease.

The early identification and characterization of patients with covid-19 in emergency units is crucial to provide the best health care, the thoracic radiography being a complementary global means of diagnosis, which is fast and easy to perform. The creation of these AI models, applied to the analysis of thoracic radiographies in patients in emergency units, assists, in combination with further clinical data, the medical professionals in the detection of suspected cases of covid-19.

The aim is to respond to the following needs:

1. Optimize the work of radiologists, identifying and prioritizing in work schedule those radiographies suspected of covid-19;
2. Assist physicians of other specialties in the context of emergency, when radiologists are not available, with a tool for the analysis of thoracic radiographies;
3. Simplify the decision making for health professionals in the definition of the clinical protocol.

This project is being developed by a multidisciplinary team of radiologists, physicians from the emergency services of Hospital da Luz Lisboa and Hospital Beatriz Ângelo, specialists in ergonomics and human factors, researchers in the areas of machine learning and artificial intelligence, specialists in information systems and managers. The project is financed under the program Portugal 2020 and results will be presented during the first semester of 2021.