Computer Aided Detection System for Pharyngitis Based on Convolutional Neural Network

Pharyngitis is inflammation of the throat or pharynx, known as strep throat, caused by viruses or bacteria. Some symptoms that arise in pharyngitis are difficulty swallowing, hoarseness of voice, itching, and sore throat. That way, a person will find it difficult to talk and eat. If there are complications in strep throat, it will trigger other diseases such as chronic tonsillitis, dengue fever, and abscesses, so it needs to be further diagnosed. This study proposed an automatic laryngitis classification method based on strep throat images to assist nose and throat specialists in diagnosing the disease. This study compares several neural network architectures with Adam optimizers, such as VGG-16, VGG-19, DenseNet121, AlexNet, and ResNet152. There are two classes of sore throat conditions classified in this study. Furthermore, a comparison is performed by augmenting the data. The simulation results show that by using non-augmented data, the proposed system achieved 88% of accuracy, while by using augmented data, the system achieved 98% accuracy. The neural network-based transfer learning method that is simulated in this study is expected to support the clinical diagnosis of strep throat.