High-level intuitive features (HLIFs) for intuitive skin lesion description.

Area of Science:

• Dermatology
• Medical Imaging
• Computer Science

Background:

• Melanoma is a deadly skin cancer with rising incidence.
• Current detection methods lack objectivity and intuitive rationale.
• Decision support systems are needed for accurate melanoma diagnosis.

Purpose of the Study:

• To propose high-level intuitive features (HLIFs) for quantitative melanoma description in standard camera images.
• To improve melanoma classification accuracy and provide intuitive diagnostic rationale.
• To address limitations of existing low-level features in high-dimensional spaces.

Main Methods:

• Development of HLIFs modeled on dermatologists' ABCD criteria.
• Integration of HLIFs with existing low-level feature sets.
• Experimental evaluation of classification accuracy and data separation capabilities.

Main Results:

• Concatenating HLIFs with low-level features significantly increased classification accuracy.
• HLIFs demonstrated superior data separation compared to low-level features.
• The proposed features enable intuitive diagnostic rationale for users.

Conclusions:

• HLIFs offer a promising approach for objective and intuitive melanoma detection.
• The developed features enhance the performance of melanoma decision support systems.
• HLIFs facilitate a better understanding of diagnostic reasoning in melanoma analysis.