Skin cancer recognition by using a neuro-fuzzy system

Area of Science:

• Dermatology
• Medical Imaging
• Artificial Intelligence

Background:

• Skin cancer is a prevalent malignancy, particularly in fair-skinned individuals, often linked to UV exposure.
• Early detection significantly improves patient outcomes by reducing mortality and morbidity.
• Current diagnostic methods can be complex, subjective, and reliant on clinician expertise.

Purpose of the Study:

• To evaluate the efficacy of image processing techniques combined with artificial intelligence for skin cancer diagnosis.
• To compare the diagnostic performance of a neural network (NN) system against a neuro-fuzzy system for classifying skin cancer types.

Main Methods:

• Utilized image processing techniques integrated with a hierarchical neural network (NN) system.
• Employed a neuro-fuzzy system, combining NN and fuzzy inference, for skin cancer detection.
• Assessed diagnostic accuracy, sensitivity, and specificity for both methodologies.

Main Results:

• The hierarchical neural network (NN) achieved a diagnostic accuracy of 90.67%.
• The neuro-fuzzy system demonstrated a slightly higher accuracy of 91.26% in diagnosing skin cancer types.
• The NN showed 95% sensitivity and 88% specificity, while the neuro-fuzzy system achieved 98% sensitivity and 89% specificity.

Conclusions:

• Image processing combined with AI, particularly neuro-fuzzy systems, offers a promising, objective approach to skin cancer diagnosis.
• Neuro-fuzzy systems show potential for improved accuracy and sensitivity in detecting various skin cancer types.
• These AI-driven methods could overcome limitations of traditional, subjective diagnostic techniques, aiding early detection efforts.