A free, open-source app developed by researchers at Stanford houses an image-based deep convolutional neural network (named MPXV-CNN) which has been shown to detect mpox lesions from photos with an accuracy of 90%. Thieme et al published information about the development and performance of the tools accessible to patients through PoxApp in Nature Medicine. Using over 138,000 non-mpox skin lesion images and nearly 700 images of skin lesions caused by the virus, researchers were able to train the neural network to accurately identify mpox lesions with good sensitivity and specificity across various skin tones and regions of the body where the lesions developed. For patients, the PoxApp is easy to use—it comprises a questionnaire and a prompt to photograph a skin lesion and takes 5 minutes or less to complete. In a companion press release from Stanford University, lead study author Alexander Thieme, MD, the lead developer of the app and a visiting scholar in the Department of Medicine from Charité–Universitätsmedizin Berlin and Berlin Institute of Health, commented, “Many people seek out medical information on the Internet, and much of that may be inaccurate…. With this app, developed with guidance from the Centers for Disease Control and Prevention and the World Health Organization, we hope to encourage people to seek out care.”
May 24, 2023