Background Interpretation of medical images needs to be standardized and trained. This applies also to fluorescence-optical imaging (FOI), a novel tool for assessment of inflammation in patients with arthritis (1). In this study we refined and evaluated the inter-reader reliability of the previously described algorithm of image interpretation of FOI.
Objectives To determine the inter-reader reliability of the fluorescence optical imaging activity score (FOIAS) 7 readers from 4 centres each evaluated 34 randomly selected cases (5 RA, 5 PsA, 5 remissions, and 19 EA). FOIAS was analyzed in PrimaVistaMode (PVM) and 3 phases (P1, P2, P3). All participants were blinded for diagnosis and clinical findings. Overall, 28560 data sets were evaluated.
Methods All participants were initially trained in software use and basic image interpretation by the supplier of the system. 5 of the 7 readers had no knowledge in the standardized image evaluation. Before starting the study a 2.5h training session was carried out and repeated once. All participants obtained a guideline and picture samples. For image interpretation, the automatically generated gain had to be checked by the readers and corrected as appropriate. The readers had to determine the 3 phases, and evaluate all 4 datasets of each case according to the guideline. Spearman correlations, agreement rates (2) and Cohen’s kappa coefficients were calculated.
Results Correlations and agreement rates are presented in table 1. Two readers (2,4) showed relevant difference to the other readers. Cohen’s Kappa was between moderate and substantial. Highest Kappa was found for PVM (κ=0.68).
Conclusions There was overall a good to nearly perfect correlation between readers. Thus standardized FOI evaluation gives high agreement in assessment of disease activity using FOIAS. There was high agreement for the decision if a joint is affected or not. Cohens Kappa showed that the standardized protocol is reliable after training. Highest Kappa was found for PVM. This is explicable with the fact, that PVM is a static image, while the image stack is dynamic. Reader 2 and 4 had problems in image interpretation depending on a dyschromatopsia and wrong image adjusting. Further training will be helpful. In conclusion, the used standardized protocol for assessment of inflammatory activity in FOI seems to be reliable and useful in daily practice. Redetermination of interreader-reliability after further refining of the protocol and training is planned.
Werner SG, Langer HE, Ohrndorf et al. Inflammation assessment in patients with arthritis using novel in vivo fluorescence optical imaging technology. Ann Rheum Dis 2011, Epub ahead of print.
Schwenke C and Busse R. Analysis of differences in proportions from clustered data with multiple measurements in diagnostic studies. Meth Inform Med 2007 (46), 548-552
Disclosure of Interest S. Werner Grant/Research support from: Pfizer, U. Käßer: None Declared, C. Amberger: None Declared, M. Piesga: None Declared, F. Spiecker: None Declared, C. Volberg: None Declared, C. Schwenke: None Declared, H.-E. Langer Grant/Research support from: Pfizer, M. Backhaus Grant/Research support from: Pfizer