Press Release
Monday, March 23, 2020
Detecting COVID-19 with Deep Learning and Lung X-Rays
Université de Moncton’s Perception, Robotics and Intelligent Machines (PRIME) group, under the supervision of Professor Moulay Akhloufi, has developed and deployed a deep neural network learning model to detect COVID-19 online from lung X-ray images.
On an X-ray, COVID-19 looks like pneumonia. The goal of Professor Akhloufi’s team was therefore to detect it without mistaking it for any other type of illness. To prevent confusion and errors, PRIME’s model was built from x-rays of persons who had contracted COVID-19, healthy persons and persons with other diseases having similar signs, such as SARS (severe acute respiratory syndrome).
“Currently, COVID-19 is detected in the field using RT-PCR (reverse transcription polymerase chain reaction) tests that are sent to specialized labs and it takes time to get the results,” Professor Akhloufi noted. “CT scan images have also been used, but scanners aren’t very widely available. X-rays are used, especially in the countries that have been most seriously affected, like Spain. That makes X-rays an attractive and widely-available tool to help diagnose the illness.”
The model developed by Professor Akhloufi’s team has a performance level of 98% (AUC) and a sensitivity of 96% (the proportion of true positives correctly identified as such).
Return to Press release