Finding COVID-19 from Chest X-rays using Deep Learning on a Small Dataset
<p>esting for COVID-19 has been unable to keep up with the demand. Further, the false negative rate is projected to be as high as 30\ and test results can take some time to obtain. X-ray machines are widely available and provide images for diagnosis quickly. This paper explores how useful chest X-ray images can be in diagnosing COVID-19 disease. We have obtained 135 chest X-rays of COVID-19 and 320 chest X-rays of viral and bacterial pneumonia. </p><p> A pre-trained deep convolutional neural network, Resnet50 was tuned on 102 COVID-19 cases and 102 other pneumonia cases in a 10-fold cross validation. The results were </p><p> an overall accuracy of 90.7% with a COVID-19 true positive rate of 0.83 and an AUC of 0.987,</p><p> Pre-trained Resnet50 and VGG16 plus our own small CNN were tuned or trained on a balanced set of COVID-19 and pneumonia chest X-rays. An ensemble of the three types of CNN classifiers was applied to a test set of 33 unseen COVID-19 and 208 pneumonia cases. The overall accuracy was 94.4% with the true positive rate for COVID-19 of 0.969 with 6% false positives for a true negative rate of 0.94 and AUC of 0.99. </p><p> </p><p> This preliminary study has flaws, most critically a lack of information about where in the disease process the COVID-19 cases were and the small data set size. More COVID-19 case images at good resolution will enable a better answer to the question of how useful chest X-rays can be for diagnosing COVID-19. </p><p> </p><p> Note an earlier version of this work inadvertently used chest X-rays of viral and bacterial pneumonia that came from a dataset of children under 5 years old and those results should be ignored. </p>