| 초록 |
Kidney pathology is generally assessed in a qualitative rather than quantitative manner from renal biopsies. In lupus nephritis, quantification requires two processes: 1) synthesising information from individual glomerular histology to define sample-level pathology, 2) estimation of errors between sample-level and organ-level pathology. Examples of pathology in this context include the ISN/RPS 2003 classification, and the diversity of glomerular histology. To facilitate the process, we developed criteria for classifying individual glomeruli into seven types, which can be used to reconstruct the ISN/RPS classification. Using our glomerular classification, we were able to obtain the proportions of
each type of glomerulus in each sample. To estimate inconsistencies between the ISN/RPS classes in a biopsy sample and the whole organ, we developed a class inconsistency rate (CIR) using bootstrapping. To quantify the diversity of glomerular histology, we developed a measure of glomerular diversity (GD) using a diversity index. The features of CIR and GD and their use for defining organ-level pathology were evaluated. CIR (15.80 ± 40.03%) increased when the number of biopsied glomeruli was low and the sample was intermediate between different classes. GD was larger in classes IV-G and V than in classes III and IV-S. Serum creatinine levels were higher in the middle-range of GD in class IV-G samples. In simulations, the GD of samples differed from those of the whole organ when low numbers of glomeruli were sampled, but converged to their true values when large numbers of glomeruli were sampled. We conclude that CIR can distinguish between cases whose classification on the ISN/RPS classification is secure from ones whose classification is less secure and GD can add to the information on the pathologic state of the kidney obtained from the ISN/RPS classification. Our procedures should be applicable to other renal pathologies. |