Biologically inspired skin lesion segmentation using a geodesic active contour technique.

Area of Science:

• Medical image analysis
• Computational dermatology
• Biomedical engineering

Background:

• Accurate skin lesion segmentation is crucial for computer-aided diagnosis of skin cancer.
• Challenges include noise from hair, skin color variations, and lighting.

Purpose of the Study:

• To develop and evaluate a novel, biologically inspired geodesic active contour (GAC) technique for robust skin lesion segmentation.
• To overcome limitations of existing methods in handling noise and achieving accurate boundary detection.

Main Methods:

• Utilized a biologically inspired geodesic active contour (GAC) algorithm for lesion segmentation.
• Implemented automatic contour initialization to avoid noise-induced artifacts.
• Incorporated biologically relevant features like spectral image subtraction and border smoothing.

Main Results:

• The GAC algorithm achieved a median XOR border error of 6.7% on 100 images, comparable to dermatologists (7.4%).
• Outperformed gradient vector flow snakes (14.2% XOR error).
• On a challenging set of 1238 low-contrast images, a median XOR error of 23.9% was obtained.

Conclusions:

• Geodesic active contour (GAC) techniques demonstrate significant potential for automatic skin lesion segmentation.
• The proposed method shows promise for improving the accuracy of computer-aided skin cancer diagnosis.