A Novel Approach to Segment Skin Lesions in Dermoscopic Images Based on a Deformable Model

• Published in: IEEE journal of biomedical and health informatics Vol. 20; no. 2; pp. 615 - 623
• Main Authors: Zhen Ma, Tavares, Joao Manuel R. S.
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.03.2016; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Clustering algorithms; Color; color spaces; Deformable models; Dermoscopy - methods; Diagnosis; Formability; Humans; Image color analysis; Image Interpretation, Computer-Assisted - methods; Image segmentation; Lesions; Level set; level set method; medical imaging; melanoma; Segmentation; Skin; Skin - diagnostic imaging; Skin Neoplasms - diagnostic imaging; Color spaces; level set method; melanoma; image segmentation; medical imaging
• ISSN: 2168-2194; 2168-2208; 2168-2208
• DOI: 10.1109/JBHI.2015.2390032

Dermoscopy is an imaging technique that has been widely used in the diagnosis of skin lesions. However, its accuracy largely depends on the dermatologist's experience; thus, computer-aided diagnosis techniques are required. In this paper, a novel approach based on a deformable model is proposed to handle the segmentation of skin lesions in dermoscopic images. The RGB color space is converted so that the color information contained in the images can be used effectively to differentiate normal skin and skin lesions; and the differences in the color channels are combined together to define the speed function and the stopping criterion of the deformable model. This novel approach is robust against the noise, and provides an effective and flexible segmentation. Two image databases were used to test the performance of the novel approach and the segmentation results obtained were satisfactory. Quantitative analysis on 250 dermoscopic images showed that the novel algorithm outperformed other state-of-the-art algorithms. Also, using comparative data, the reliability and the implementation issues of the approach are discussed in this paper.