Automated detection of COVID-19 in lung CTs using comprehensive patient information
During earlier phases of the COVID-19 pandemic, the emphasis was on fast, reliable identification of COVID-19 patients in screening settings. Now the focus is shifting towards differential diagnosis and decision support for treatment and patient management, including:
(1) Differential diagnosis and assessment of suspected COVD-19 patients for whom PCR testing was negative
(2) Prediction of disease course for individuals, timely treatment decisions and patient management.
contextflow’s current software identifies COVID-19-related disease patterns in lung CTs and provides quantitative measurements to support radiologists during the diagnostic process. The next phase enabled by this project will develop this technology further to include the prediction of severity, disease course and need for intensive care or mechanical ventilation in COVID-19 patients, which will be increasingly important in the next phase of the pandemic. This is achieved by building on the current contextflow SEARCH and TRIAGE products by including additional anonymised patient information into the machine learning models, and hence enabling radiologists to be alerted to the potential presence of signs of a COVID-19 infection in CT scans of the lungs.