Transformer-based decoder of melanoma classification using hand-crafted texture feature fusion and Gray Wolf...

Area of Science:

• Dermatology
• Medical Imaging
• Artificial Intelligence

Background:

• Melanoma detection is crucial for effective treatment, necessitating accurate and efficient diagnostic tools.
• Current methods may lack the precision required for early-stage melanoma identification.
• Integrating advanced computational techniques can improve skin cancer classification accuracy.

Purpose of the Study:

• To develop and validate a novel approach for enhanced melanoma classification.
• To improve the efficiency and accuracy of melanoma detection using a hybrid AI model.
• To leverage texture analysis and optimization algorithms within a transformer framework.

Main Methods:

• Image preprocessing including median filtering for quality enhancement.
• Extraction of texture features using Gray-Level Co-occurrence Matrix (GLCM) and Local Binary Patterns (LBP).
• Feature selection via Gray Wolf Optimization (GWO) and classification using a transformer-based decoder.

Main Results:

• The proposed method achieved high accuracy (99.54% on HAM10000, 99.47% on ISIC2019) and F1-scores (99.11% on HAM10000, 99.25% on ISIC2019).
• The integration of hand-crafted texture features with a transformer model proved effective for melanoma classification.
• Experimental validation on benchmark datasets confirmed the methodology's superior performance.

Conclusions:

• The developed transformer-based model with GWO-selected texture features offers a highly effective solution for melanoma detection.
• This approach demonstrates significant potential for improving diagnostic accuracy in clinical settings.
• The study highlights the synergy between classical image features and modern deep learning architectures for medical image analysis.