Automatic Classification of Specific Melanocytic Lesions Using Artificial Intelligence.

Area of Science:

• Dermatology
• Medical Imaging
• Artificial Intelligence

Background:

• Early melanoma detection is critical due to metastasis and limited treatment options for advanced cases.
• Computer-aided diagnosis (CAD) systems aim to enhance melanoma detection sensitivity and specificity.
• A system for classifying diverse melanocytic lesion types, beyond benign/malignant, was previously unavailable.

Purpose of the Study:

• To introduce a novel approach for classifying various melanocytic lesions.
• To develop an automated algorithm capable of differentiating between melanoma, Clark nevus, Spitz/Reed nevus, and blue nevus.
• To advance computer-aided diagnosis for more precise skin lesion categorization.

Main Methods:

• The proposed algorithm incorporates image enhancement, lesion segmentation, and feature extraction/selection.
• A classification model is applied to categorize the segmented and processed skin lesion images.
• The methodology focuses on a comprehensive analysis of dermoscopic images.

Main Results:

• The algorithm achieved a 92% accuracy rate in classifying melanocytic lesions.
• Testing was conducted on a dataset of 300 dermoscopic images.
• The results demonstrate high correctness in classifying various types of melanocytic lesions.

Conclusions:

• The developed system aids in precise diagnosis of skin mole types.
• Implementation of this system can lead to a reduction in the number of biopsies performed.
• The technology has the potential to decrease the morbidity associated with skin lesion excisions.