20 March 2014 Registration and color calibration for dermoscopy images in time-course analysis
Author Affiliations +
Abstract
Since melanomas grow and metastasize rapidly, the mutation in their appearance is much larger than that of nevi. If the variation of skin tumor can be evaluated quantitatively, it is of substantial help not only for clinical diagnosis, but also for development of computer-based diagnostic systems. However, photographic conditions of skin tumor are in most cases not uniform during the follow-up. In this study, we proposed a fully automated image registration and color calibration method between dermoscopy images in the time-course analysis. Our proposed algorithm aligned the time-course images with a precision of 91.6 ± 5.1% and a recall of 95.7 ± 5.9%, respectively whereas the fully manual registrations with Exif data as a performance reference did 95.4 ± 3.2% and 92.4 ± 6.5%, respectively. Our color calibration method largely reduced the color difference between timecourse images ΔE from 10.9 ± 5.6 to 3.9 1.7. These results showed that the proposed method was effective to compensate both geometrical and chronological changes between dermoscopy images in the time-course analysis.
© (2014) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Daiji Furusho, Hitoshi Iyatomi, "Registration and color calibration for dermoscopy images in time-course analysis", Proc. SPIE 9035, Medical Imaging 2014: Computer-Aided Diagnosis, 90353B (20 March 2014); doi: 10.1117/12.2044019; https://doi.org/10.1117/12.2044019
PROCEEDINGS
6 PAGES


SHARE
Back to Top