Detection of monkeypox skin lesions using edge enhancement algorithms integrated with hybrid deep learning

Area of Science:

• Artificial Intelligence in Dermatology
• Medical Image Analysis
• Infectious Disease Diagnostics

Background:

• Monkeypox poses a global health threat, necessitating advanced detection methods beyond traditional polymerase chain reaction (PCR).
• PCR is limited by cost, time, and contagion risks to healthcare professionals.
• Innovative solutions are crucial for efficient and safe monkeypox diagnosis.

Purpose of the Study:

• To introduce a lightweight deep learning framework for enhanced monkeypox lesion detection using skin images.
• To improve the accuracy and efficiency of monkeypox diagnosis through AI.

Main Methods:

• A novel edge enhancement algorithm incorporating contrast-limited adaptive histogram equalization and bilateral filters was applied to refine skin images.
• The framework was evaluated using six pre-trained deep learning models and a hybrid model (DenseNet121 + ConvNeXt-Tiny, DN-CXT).
• Performance metrics included accuracy, F1-score, and precision, with optimization via Adam, RMSprop, and SGD.

Main Results:

• The proposed DN-CXT model demonstrated superior performance with 97% test accuracy, 97% F1-score, and 99% precision.
• Other models like DenseNet121, MobileNetV2, InceptionV3, and ConvNeXt-Tiny also yielded high performance.
• The framework effectively enhances monkeypox lesion detection from skin images.

Conclusions:

• The developed deep learning framework significantly advances medical image analysis for monkeypox lesion detection.
• AI-driven methodologies can be integrated into monkeypox detection workflows for improved diagnostic efficiency.
• This approach enhances healthcare response to emerging infectious diseases and reduces risks to medical personnel.