Content-Based Medical Image Retrieval System for Skin Melanoma Diagnosis Based on Optimized Pair-Wise Comparison...

Area of Science:

• Dermatology
• Medical Imaging
• Artificial Intelligence

Background:

• Accurate diagnosis of skin conditions, particularly melanoma, is crucial to prevent delayed or incorrect treatment.
• Physicians may miss subtle or atypical lesions, highlighting the need for diagnostic support systems.
• Leveraging healthcare databases for similar case studies can improve diagnostic accuracy.

Purpose of the Study:

• To develop an assistive system for dermatologists to accurately identify 23 different types of melanoma.
• To enhance the accuracy of skin-melanoma image search systems through advanced feature weighting techniques.

Main Methods:

• Training a skin-melanoma similar image search system with 2300 dermoscopic images.
• Implementing a novel optimized pair-wise comparison (OPWC) approach for dynamic feature weighting.
• Utilizing analytic hierarchy process (AHP) and meta-heuristic optimization algorithms (PSO, JAYA, GA, GWO) to optimize feature weights.

Main Results:

• The proposed system demonstrated significant precision and recall in identifying 23 classes of melanoma.
• Dynamic weighting of low-level features based on image characteristics improved search accuracy.
• The OPWC approach proved effective in assigning optimal weights for melanoma classification.

Conclusions:

• The developed assistive system can significantly aid dermatologists in the accurate identification of various melanoma types.
• This AI-driven approach offers a valuable tool for expert decision support in clinical dermatology.
• The dynamic feature weighting method represents a novel advancement in medical image analysis for skin cancer detection.