Agreement between mathematical methods and raters in sample subsets.

The Fleiss' κ values obtained for selected subsets are stratified according to the raters one another or the raters together with either the CT or MR methods. The query subset was obtained by considering as negative the CT results with a Cq above 40 cycles; the other subsets included also Cq above this level to be consistent with the original clinical dataset. The p-values for the Fleiss' κ test were all lower than 0.001. The accuracy was based on the consensus classification. The number of reactions per group is reported (n).