A multi-institution research team has developed an artificial intelligence (AI) model that can rapidly detect COVID-19 in chest X-rays with greater than 98% accuracy. The deep-learning application was able to differentiate not only between healthy lungs and those infected with COVID-19, but it can also diagnose patients with all-cause pneumonia solely from an X-ray image. The results are published in Scientific Reports. [1]
“The most widely used COVID-19 test, real time polymerase chain reaction (PCR), can be slow and costly, and produce false-negatives. To confirm a diagnosis, radiologists need to manually examine CT scans or X-rays, which can be time consuming and prone to error,” study co-author Amir Gandomi from Australia’s University of Technology Sydney said in a statement announcing the study results.
Given the time-consuming nature of having a radiologist review X-Ray images—combined with the potential for error—the research team sought to measure how well AI can speed up the process. For their study, they developed a deep-learning program from the popular convolutional neural network (CNN) model, which typically performs well analyzing images and has been used in other studies involving X-rays.
The custom-CNN model was trained on a dataset of images of normal lungs, those of COVID-19 patients, and X-rays of lungs where viral pneumonia was present. Combination images were also used, such as lungs that showed COVID-19 infection without pneumonia, and images from patients who had pneumonia but not COVID-19.
When reviewing images outside its training data, the model showed a greater than 98% accuracy of diagnosing a patient based on the lung X-rays. When the custom CNN was pitted against other AI, it outperformed them as well, signaling that the custom-CNN developed by the researchers may have real-world value for the diagnosis of COVID-19.
“The new AI system could be particularly beneficial in countries experiencing high levels of COVID-19 where there is a shortage of radiologists. Chest X-rays are portable, widely available and provide lower exposure to ionizing radiation than CT scans,” Gandomi concluded.
Chad is an award-winning writer and editor with over 15 years of experience working in media. He has a decade-long professional background in healthcare, working as a writer and in public relations.