Convolutional Neural Network for Pharyngitis Classification
Deep Learning Model and Application for the Diagnosis of Exudative Pharyngitis
Principal Investigator: Dr Cheng Tim-Ee Lionel, Clinical Director (Artificial Intelligence), Future Health System Department, Singapore General Hospital
Introduction
Telemedicine is firmly established in the healthcare landscape of many countries. Acute respiratory infection is the most common reason for a telemedicine consultation. Throat examination is important to diagnose bacteria pharyngitis, but this is challenging for doctors during a telemedicine consultation. A solution is for patients to upload images of their throat to a web application.
Objective
The aim is to develop a deep learning model for the automated diagnosis of exudative pharyngitis. Thereafter, the model will be deployed online.
Methods
We used 343 throat images (139 with exudative pharyngitis and 204 without pharyngitis) in the study. ImageDataGenerator augmented the training data. Transfer learning (pre-trained on ImageNet data) was used. The convolutional neural network models of MobileNetV3, ResNet50 and EfficientNetB0 were implemented to train the dataset, with hyper-parameter tuning.
Results
All 3 models trained successfully with loss and training loss decreasing, and accuracy and training accuracy increasing with successive epochs. The EfficientNetB0 model achieved the highest accuracy at 95.5%, compared to MobileNetV3 (82.1%) and ResNet50 (88.1%). The EfficientNetB0 model was also able to achieve a high precision (1.00), recall (0.89) and f1-score (0.94).
Discussion
We have trained a deep learning model based on EfficientNetB0 that can diagnose exudative pharyngitis. Our model was able to achieve the highest accuracy at 95.5% out of all previous studies that used machine learning for the diagnosis of exudative pharyngitis. We have deployed the model on a web application that can be used to augment the doctor’s diagnosis of exudative pharyngitis.
