Background: For optimizing and evaluating image quality in medical imaging, one can use visual grading\nexperiments, where observers rate some aspect of image quality on an ordinal scale. To analyze the grading\ndata, several regression methods are available, and this study aimed at empirically comparing such techniques,\nin particular when including random effects in the models, which is appropriate for observers and patients.\nMethods: Data were taken from a previous study where 6 observers graded or ranked in 40 patients the image\nquality of four imaging protocols, differing in radiation dose and image reconstruction method. The models tested\nincluded linear regression, the proportional odds model for ordinal logistic regression, the partial proportional odds\nmodel, the stereotype logistic regression model and rank-order logistic regression (for ranking data). In the first two\nmodels, random effects as well as fixed effects could be included; in the remaining three, only fixed effects.\nResults: In general, the goodness of fit (AIC and McFadden�s Pseudo R2) showed small differences between the\nmodels with fixed effects only. For the mixed-effects models, higher AIC and lower Pseudo R2 was obtained, which\nmay be related to the different number of parameters in these models. The estimated potential for dose reduction\nby new image reconstruction methods varied only slightly between models.\nConclusions: The authors suggest that the most suitable approach may be to use ordinal logistic regression, which\ncan handle ordinal data and random effects appropriately.
Loading....