Email updates

Keep up to date with the latest news and content from BMC Bioinformatics and BioMed Central.

This article is part of the supplement: Proceedings of the Eighth Annual MCBIOS Conference. Computational Biology and Bioinformatics for a New Decade

Open Access Proceedings

An improved border detection in dermoscopy images for density based clustering

Sait Suer1, Sinan Kockara1* and Mutlu Mete2

Author Affiliations

1 Computer Science, University of Central Arkansas, 201 Donaghey Ave. Conway, 72035, AR, USA

2 Computer Science and Information System, Texas A&M University – Commerce, 75429, TX, USA

For all author emails, please log on.

BMC Bioinformatics 2011, 12(Suppl 10):S12  doi:10.1186/1471-2105-12-S10-S12

Published: 18 October 2011

Abstract

Background

Dermoscopy is one of the major imaging modalities used in the diagnosis of melanoma and other pigmented skin lesions. In current practice, dermatologists determine lesion area by manually drawing lesion borders. Therefore, automated assessment tools for dermoscopy images have become an important research field mainly because of inter- and intra-observer variations in human interpretation. One of the most important steps in dermoscopy image analysis is automated detection of lesion borders. To our knowledge, in our 2010 study we achieved one of the highest accuracy rates in the automated lesion border detection field by using modified density based clustering algorithm. In the previous study, we proposed a novel method which removes redundant computations in well-known spatial density based clustering algorithm, DBSCAN; thus, in turn it speeds up clustering process considerably.

Findings

Our previous study was heavily dependent on the pre-processing step which creates a binary image from original image. In this study, we embed a new distance measure to the existing algorithm. This provides twofold benefits. First, since new approach removes pre-processing step, it directly works on color images instead of binary ones. Thus, very important color information is not lost. Second, accuracy of delineated lesion borders is improved on 75% of 100 dermoscopy image dataset.

Conclusion

Previous and improved methods are tested within the same dermoscopy dataset along with the same set of dermatologist drawn ground truth images. Results revealed that the improved method directly works on color images without any pre-processing and generates more accurate results than existing method.