Hyper-parameter tuned deep learning approach for effective human monkeypox disease detection.

Area of Science:

• Medical Informatics
• Computer Vision
• Epidemiology

Background:

• Human monkeypox is a zoonotic viral disease with potential for societal devastation.
• Early diagnosis is crucial for effective management and containment of monkeypox outbreaks.
• Limitations in Polymerase Chain Reaction (PCR) testing availability necessitate alternative diagnostic approaches.

Purpose of the Study:

• To develop and evaluate an accurate and resilient deep learning model for human monkeypox detection.
• To differentiate monkeypox lesions from other skin conditions, such as chickenpox, using medical images.
• To explore the efficacy of hyperparameter optimization techniques for enhancing model performance.

Main Methods:

• Utilized a combination of convolutional neural networks and transfer learning for feature extraction from medical images.
• Implemented hyperparameter optimization strategies, including SDG optimizer, Bayesian optimizer, and Learning without Forgetting.
• Developed a Yolov5 model for classifying skin lesions, trained on the Roboflow skin lesion dataset.

Main Results:

• The proposed deep learning model achieved a high classification accuracy of 98.18% for monkeypox skin lesions.
• The model demonstrated superior performance compared to existing state-of-the-art models.
• The hyperparameter tuning strategies significantly improved the model's accuracy and resilience.

Conclusions:

• The developed deep learning model shows significant promise for accurate and efficient human monkeypox detection.
• This computer-assisted approach can serve as a valuable tool in clinical settings, especially where PCR testing is limited.
• Further validation in real-world clinical scenarios is recommended to confirm its utility in disease diagnosis.