Comparison between manual and semi-automated radiographic scoring in rheumatoid arthritis

Y. M. J. Lim, J. S. Gregory, R. J. Barr, N. Basu, K. Secombes, D. M. Reid

Research output: Contribution to journalAbstractpeer-review


Background: Rheumatoid arthritis is an inflammatory condition that results in joint damage. Sharp-van der Heidje is one of the most widely used scoring method in assessing damage in RA. The aim of this project was to compare scoring of hand radiographs manually with a semi-automated workflow tool in development.
Method and results: Twenty-one sets of bilateral hand radiographs from patients with RA of variable duration were scored both manually and with the semi-automated method in a randomised order twice, with the time taken for scoring recorded. A paired T-test for timing showed the semi-automated method to be faster by 2.6 and 1.4 min in both the first and second runs, respectively (P = 0.0019 and P = 0.0003). Bland and Altman testing showed only one outlier with the semi-automated method suggesting a good agreement between runs with insignificant loss in repeatability. However, the new method did not improve reproducibility with coefficient of variation of 21.05% and 17.14% for the two runs compared to manual scoring (14.64% and 16.44%). Concordance coefficient testing showed good repeatability for both methods for both erosions and joint space narrowing.
Conclusion: The new semi-automated scoring system has value in terms of speed of scoring radiographs with negligible loss in accuracy, precision, and reproducibility as compared to normal manual scoring. With further development, the method has the potential to be a valuable tool in improving efficiency of assessing hand radiographs.
Original languageEnglish
Pages (from-to)E30-E31
Number of pages2
JournalScottish Medical Journal
Issue number3
Publication statusPublished - Aug 2015


Dive into the research topics of 'Comparison between manual and semi-automated radiographic scoring in rheumatoid arthritis'. Together they form a unique fingerprint.

Cite this