Image-Based Detection of Herpes Zoster Using Deep Learning

Herpes Zoster, or shingles, is an infection caused by reactivation of the Varicella-Zoster Virus. Early detection is vital to avoid this infection. This project helps us understand how one can use deep learning for automatic Herpes Zoster diagnosis from medical images. Lightweight convolutional neural network models like MobileNetV2 and VGG16 (optional) were used in combination with basic image pre-processing. For comparison, traditional classifiers like Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) were also applied. The proposed system has high accuracy with small data. To provide results in a simpler form, Gradient-weighted Class Activation Mapping (Grad-CAM) was used to create heatmap visualizations showing the affected area and ensuring the models focus on relevant parts. Results show that even using simple models and small datasets, we can identify early-stage HZ accurately, which can be a cost-effective tool to aid healthcare professionals.