Transfer learning for segmentation with hybrid classification to Detect Melanoma Skin Cancer.

Area of Science:

• Dermatology
• Medical Imaging
• Computer Science

Background:

• Melanoma is a common skin cancer often linked to sun exposure, but can occur elsewhere.
• Accurate segmentation and classification of melanoma are crucial for effective treatment.
• Biomedical image analysis offers promising tools for analyzing melanoma tissues.

Purpose of the Study:

• To address the challenges in melanoma skin cancer segmentation and classification.
• To evaluate the effectiveness of an Attribute Selected Classifier with a Color Layout Filter for melanoma detection.
• To improve diagnostic accuracy for melanoma skin cancer.

Main Methods:

• The study employed an Attribute Selected Classifier integrated with a Color Layout Filter for image enhancement.
• Segmentation and classification of melanoma skin lesions were performed using this model.
• Performance was evaluated using standard metrics like accuracy, precision, and recall.

Main Results:

• The proposed method achieved high performance metrics: 90.96% accuracy, 91% precision, and 0.91 recall.
• Further results included 0.95 ROC AUC, 0.87 Kappa Statistic, and 0.91 F-Measure.
• The research noted minimal errors with the implemented method on the tested dataset.

Conclusions:

• Attribute selection combined with image enhancement techniques like the Color Layout Filter is an efficient approach for melanoma classification.
• The proposed method demonstrates significant potential for improving the accuracy of melanoma detection in biomedical imaging.
• Further research is recommended to validate these findings on diverse datasets.