Toward automated severe pharyngitis detection with smartphone camera using deep learning networks.

摘要: Abstract Purpose Severe pharyngitis is frequently associated with inflammations caused by streptococcal pharyngitis, which can cause immune-mediated and post-infectious complications. The recent global pandemic of coronavirus disease (COVID-19) encourages the use telemedicine for patients respiratory symptoms. This study, therefore, purposes automated detection severe using a deep learning framework self-taken throat images. Methods A dataset composed two classes 131 images 208 normal was collected. Before training classifier, we constructed cycle consistency generative adversarial network (CycleGAN) to augment dataset. ResNet50, Inception-v3, MobileNet-v2 architectures were trained transfer validated randomly selected test performance models evaluated based on accuracy area under receiver operating characteristic curve (ROC-AUC). Results CycleGAN-based synthetic reflected pragmatic features pharyngitis. Using images, model demonstrated significant improvement in diagnosis. ResNet50 GAN-based augmentation showed best ROC-AUC 0.988 In 4-fold cross-validation highest achieved 95.3% 0.992, respectively. Conclusion smartphone-based screening allows fast identification potential timely diagnosis COVID-19, this will help upper symptoms improve convenience reduce transmission.