Detection of COVID-19 by an integrated system of a portable thermal camera and machine learning algorithm
Thermography is a non-ionizing, non-invasive, and low-cost imaging modality. Skin temperature can reflect the presence of inflammation in underlying tissues or where blood flow is increased or decreased because of a clinical abnormality. We developed a novel integrated image processing and machine learning algorithm for COVID-19 detection. Following IRB approval, we captured thermal images of the back of individuals with and without COVID-19 using a portable thermal camera that is connected directly to smartphones. Our algorithm computed and stored several texture parameters obtained from the thermal images. The next step was to test the ability of the algorithm to detect COVID-19. We used a linear SVM classifier with 5-fold cross-validation. The obtained sensitivity was 95%, and the specificity was 71%. In summary, we show that an integrated system of portable thermal camera and machine learning algorithm can be used to detect COVID-19.Non-contact thermal infrared imaging is a simple
Oshrit A. Hoffer
Afeka College of Engineering (Israel)
Dr. Oshrit Hoffer is a lecturer in the Department of Electrical Engineering at Afeka College of Engineering in Tel Aviv Israel. Her areas of expertise are thermal imaging, image processing and machine learning.