Quantitative color analysis for capillaroscopy image segmentation

• Published in: Medical & biological engineering & computing Vol. 50; no. 6; pp. 567 - 574
• Main Authors: Goffredo, Michela, Schmid, Maurizio, Conforto, Silvia, Amorosi, Beatrice, D’Alessio, Tommaso, Palma, Claudio
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer-Verlag 01.06.2012; Springer Nature B.V
• Subjects: Algorithms; Annotations; Biomedical and Life Sciences; Biomedical Engineering and Bioengineering; Biomedical research; Biomedicine; Blood vessels; Capillaries - pathology; Capillarity; Channels; Color; Computer Applications; Decomposition; Digital; Human Physiology; Humans; Image contrast; Image Interpretation, Computer-Assisted - methods; Imaging; Medical imaging; Microscopic Angioscopy - methods; Morphology; Nails - blood supply; Original Article; Radiology; Rheumatic diseases; Segmentation; Studies; Italy; Color analysis; Segmentation; Capillaroscopy; Color space
• ISSN: 0140-0118; 1741-0444; 1741-0444
• DOI: 10.1007/s11517-012-0907-7

This communication introduces a novel approach for quantitatively evaluating the role of color space decomposition in digital nailfold capillaroscopy analysis. It is clinically recognized that any alterations of the capillary pattern, at the periungual skin region, are directly related to dermatologic and rheumatic diseases. The proposed algorithm for the segmentation of digital capillaroscopy images is optimized with respect to the choice of the color space and the contrast variation. Since the color space is a critical factor for segmenting low-contrast images, an exhaustive comparison between different color channels is conducted and a novel color channel combination is presented. Results from images of 15 healthy subjects are compared with annotated data, i.e. selected images approved by clinicians. By comparison, a set of figures of merit, which highlights the algorithm capability to correctly segment capillaries, their shape and their number, is extracted. Experimental tests depict that the optimized procedure for capillaries segmentation, based on a novel color channel combination, presents values of average accuracy higher than 0.8, and extracts capillaries whose shape and granularity are acceptable. The obtained results are particularly encouraging for future developments on the classification of capillary patterns with respect to dermatologic and rheumatic diseases.