Biological indexes based reflectional asymmetry for classifying cutaneous lesions

Area of Science:

• Dermatology
• Medical Imaging
• Computer Vision

Background:

• Cutaneous lesions require accurate discrimination between malignant melanoma and benign nevi.
• Quantifying lesion asymmetry is crucial for diagnosis.
• Existing methods may lack precision in asymmetry assessment.

Purpose of the Study:

• To propose a novel reflectional asymmetry descriptor for quantifying cutaneous lesion asymmetry.
• To enhance the discrimination between malignant melanoma and benign nevi.
• To utilize biological indexes for improved characterization of pigmentation.

Main Methods:

• Constructing a pigmentation elevation model using biological indexes (Melanin and Erythema Indexes).
• Computing a reflectional asymmetry descriptor by minimizing histogram differences of global point signatures.
• Utilizing 311 dermoscopy images for algorithm validation.

Main Results:

• Achieved 88.50% sensitivity in discriminating malignant melanoma from benign nevi.
• Obtained 81.92% specificity using a Support Vector Machine (SVM) classifier.
• Demonstrated the effectiveness of Melanin and Erythema Indexes in characterizing pigmentation distribution.

Conclusions:

• The proposed reflectional asymmetry descriptor effectively quantifies lesion asymmetry.
• The method shows significant potential for improving the diagnostic accuracy of melanoma detection.
• The use of biological indexes offers a robust approach to analyzing cutaneous lesion characteristics.