Automatic segmentation of dermoscopy images using saliency combined with Otsu threshold

Area of Science:

• Medical Image Analysis
• Computer-Aided Diagnosis
• Dermatology

Background:

• Accurate skin cancer segmentation in dermoscopy images is vital for computer-aided diagnosis (CAD).
• Existing methods may struggle with precise lesion border extraction.

Purpose of the Study:

• To develop a novel automatic segmentation algorithm for enhanced accuracy in skin lesion border detection.
• To improve the performance of computer-aided diagnosis systems for skin cancer.

Main Methods:

• Proposed a two-stage algorithm: image enhancement and lesion segmentation.
• Image enhancement involved fusing color and brightness saliency maps derived from healthy skin priors.
• Segmentation utilized an optimized Otsu threshold method based on the enhanced image histogram.

Main Results:

• The enhancement stage effectively improved image quality by fusing saliency maps.
• The optimized Otsu threshold provided more accurate lesion borders compared to traditional methods.
• Experimental validation confirmed the robustness and superior performance of the proposed model.

Conclusions:

• The novel automatic segmentation algorithm demonstrates significant improvements in enhancement effectiveness and segmentation accuracy.
• This method offers a robust and high-performing solution for skin cancer segmentation in dermoscopy.
• The approach shows potential for advancing computer-aided diagnosis in dermatology.