Deep Learning Model and its Application for the Diagnosis of Exudative Pharyngitis

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Deep learning for disease diagnosis

Background:

• Telemedicine is widely used for acute respiratory infections.
• Throat examination is crucial for diagnosing bacterial pharyngitis but difficult remotely.
• Automated diagnosis using patient-submitted images offers a potential solution.

Purpose of the Study:

• To develop a deep learning model for automated exudative pharyngitis diagnosis.
• To deploy the diagnostic model via a web application for clinical use.

Main Methods:

• Trained convolutional neural network models (MobileNetV3, ResNet50, EfficientNetB0) on 343 throat images.
• Utilized ImageDataGenerator for data augmentation.
• Performed hyperparameter tuning for model optimization.

Main Results:

• EfficientNetB0 achieved the highest accuracy at 95.5%.
• MobileNetV3 and ResNet50 showed accuracies of 82.1% and 88.1%, respectively.
• The EfficientNetB0 model demonstrated high precision (1.00), recall (0.89), and F1-score (0.94).

Conclusions:

• A deep learning model based on EfficientNetB0 effectively diagnoses exudative pharyngitis.
• The model's 95.5% accuracy surpasses previous machine learning studies.
• The deployed web application can support clinical diagnosis of exudative pharyngitis.