Skin Diseases Classification Using Hybrid AI Based Localization Approach.

Area of Science:

• Dermatology and Artificial Intelligence
• Medical Imaging Analysis
• Computational Pathology

Background:

• Skin cancer diagnosis relies on visual inspection and dermoscopy, with early detection crucial for patient outcomes.
• Deep learning algorithms analyzing annotated skin images show promise for skin lesion classification.
• Accurate disease categorization remains a challenge despite various identification strategies.

Purpose of the Study:

• To enhance the accuracy of skin cancer detection and classification using advanced computational methods.
• To introduce a novel feature extraction technique for improved diagnostic capabilities.
• To evaluate the performance of an enhanced convolutional neural network in identifying skin lesions.

Main Methods:

• Feature extraction using Spectral Centroid Magnitude (SCM).
• Application of a median filter during the initial preprocessing stage.
• Classification of the dataset utilizing an enhanced convolutional neural network (CNN).

Main Results:

• The developed SCM method and enhanced CNN demonstrated improved outcomes in skin lesion classification.
• Comparison with current methods indicated enhanced accuracy in disease detection.
• The preprocessing steps, including median filtering, contributed to the overall performance.

Conclusions:

• The integration of SCM and enhanced CNN offers a robust approach for early and accurate skin cancer diagnosis.
• Computer-aided diagnosis, though underutilized, is vital for advancing dermatological diagnostics.
• Further research and implementation of these AI-driven methods can significantly improve patient care in dermatology.