Ai Based Diagnosis Of Covid-19 From Ct/X-ray Imaging (AICovIX+)

The gold standard for Covid-19 diagnosis is RT-PCR for SARS-CoV-2 from the upper airways. This method suffers from decreased accuracy in more severe disease stages affecting the lower airways. False negative PCR tests in severe cases impose a massive risk to the health system, promoting intra-hospital disease spread in the actual pandemic situation. Reports from China indicate false negative PCR from upper airway samples in severe Covid-19 pneumonia in more than 50% of the cases. Therefore, complementary tests for the detection of these dangerous cases are urgently needed. The value of chest CT and/or X-ray has been demonstrated as complementary diagnostics. Luxembourg has already reacted to these needs, and is urgently acquiring four additional CT scanners only for lung CTs. Based on our experience in deep learning for medical imaging our project aims to rapidly provide AI tools to speed-up diagnosis of high-risk cases from medical imaging. Our short-term objective is a an AI tool for fast and accurate discrimination between Covid-19, other pneumonia, or healthy. This information can be key for early treatment as well as for the safety of health workers, hospitals and other patients. A basic version of an AI tool is already in training and aims to be applied as soon as possible during the current pandemic. Additionally, long term research and preparation to future scenarios is planned and will be integrated with diagnostic approaches using other biomarkers for disease staging by aligning with other groups in Luxembourg.