Multi-scale classification based lesion segmentation for dermoscopic images.

Area of Science:

• Dermatology
• Medical Image Analysis
• Computer Vision

Background:

• Accurate segmentation of skin lesions is crucial for diagnosis.
• Existing methods face challenges in handling variations in lesion appearance and image quality.

Purpose of the Study:

• To develop a robust and accurate lesion segmentation method for dermoscopic images.
• To improve the identification of lesion boundaries compared to current techniques.

Main Methods:

• A multi-scale classification approach using multiple classifiers trained at various resolutions.
• Pixel-level fusion of classifier outputs to create saliency maps.
• Otsu thresholding applied to saliency maps for binary mask generation.

Main Results:

• The proposed method achieved a Dice Coefficient of 0.91 and an accuracy of 94% on two public datasets.
• Demonstrated superior performance compared to existing lesion segmentation methods.
• Effective in identifying lesion boundaries in dermoscopic images.

Conclusions:

• The multi-scale classification method offers a robust solution for skin lesion segmentation.
• The technique shows significant potential for improving automated skin lesion analysis.
• High accuracy and Dice Coefficient indicate clinical relevance for diagnostic support.