Deep learning-based skin lesion analysis using hybrid ResUNet++ and modified AlexNet-Random Forest for enhanced...

Area of Science:

• Dermatology and Medical Imaging
• Artificial Intelligence in Healthcare
• Computational Pathology

Background:

• Skin cancer is a leading cause of mortality globally, with early detection critical for patient survival.
• Accurate diagnosis and treatment of skin lesions are hampered by challenges in segmentation and classification.
• Current diagnostic methods require enhancement for improved precision and efficiency.

Purpose of the Study:

• To develop and validate a novel hybrid deep learning model for enhanced segmentation and classification of skin lesions.
• To improve the accuracy of distinguishing between benign and malignant skin lesions.
• To establish a more effective approach for early-stage skin cancer detection and management.

Main Methods:

• A hybrid deep learning approach combining ResUNet++ for segmentation and a modified AlexNet-Random Forest (AlexNet-RF) for classification.
• Preprocessing techniques, including morphology-based hair removal, to improve segmentation accuracy.
• Intensive validation of the model on the widely-used Ham10000 dataset for skin lesion analysis.

Main Results:

• The proposed hybrid model demonstrated superior performance in both skin lesion segmentation and classification compared to existing methods.
• ResUNet++ achieved accurate lesion segmentation, while the AlexNet-RF classifier provided robust and precise classification.
• The model effectively isolated lesions, reducing background noise and improving the analysis of anatomical features.

Conclusions:

• The hybrid deep learning model offers a significant advancement in skin lesion analysis, leading to more accurate diagnoses.
• This approach holds promise for improving early detection rates of skin cancer, potentially saving lives.
• The combination of ResUNet++ and AlexNet-RF provides a powerful tool for dermatological applications.