Epidermis lesion detection via optimized distributed capsule neural network.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Skin cancer, including melanoma, poses a significant global health risk, with early diagnosis critical for survival.
• Accurate diagnosis of skin lesions remains challenging due to symptom complexity and variability.
• Existing diagnostic methods require enhancement for improved accuracy and efficiency.

Purpose of the Study:

• To introduce a novel computational approach for enhanced skin cancer detection.
• To improve the accuracy and reliability of diagnosing skin lesions, particularly melanoma.
• To develop an optimized deep learning model for precise epidermis lesion identification.

Main Methods:

• A Golden Hawk Optimization-based Distributed Capsule Neural Network (GHO-DCaNN) was developed.
• An optimized clustering-based segmentation approach using Sewer Shad Fly Optimization (SSFO) was integrated.
• The GHO-DCaNN model was trained using a Hybrid GHO optimizer inspired by golden eagle and fire hawk behaviors.

Main Results:

• The GHO-DCaNN achieved high performance metrics for skin lesion detection.
• Specificity, sensitivity, and accuracy rates reached up to 97.53%, 99.05%, and 98.83% with 90% training data.
• With 10-fold cross-validation, performance improved to 97.83% specificity, 99.50% sensitivity, and 99.06% accuracy.

Conclusions:

• The proposed GHO-DCaNN demonstrates significant potential for accurate and efficient skin cancer diagnosis.
• The integration of advanced optimization algorithms enhances lesion segmentation and feature extraction.
• This computational approach offers a promising advancement in the field of medical imaging for dermatology.