[1699] Efficient Methods for Morphometric Analysis of Cortical Intersitital Volume Fraction in Protocol Kidney Transplant Biopsies
Shimi Sharief, Suman Setty, Sanjeev Akkina. Rush University, Chicago, IL; University of Illinois at Chicago, Chicago, IL
Background: The use of the Banff criteria for scoring interstitial fibrosis in clinically stable kidney transplant recipients is problematic due to low degrees of fibrosis usually seen in protocol biopsies. For clinical studies, morphometric approaches such as point counting provide a more precise measure of fibrosis but at the expense of time. Our objective was to compare point counting with two computer-based methods of measurement of cortical interstitial volume fraction (Vvint).
Design: Ten randomly chosen protocol biopsies stained with Masson trichrome were scanned at high-resolution and then divided into individual images at 200X for quantitative analysis. Point counting was done using ImageJ software (NIH.gov) with a grid of 768 points. The point count within the interstitium excluding the Bowman's capsule, tubular basement membranes, and blood vessels was enumerated and divided by the total number of points in the cortex to give the Vvint. For computer-based tools, we used ImageJ with the color deconvolution plugin and Histolab (Microvision Instruments, France). The ImageJ color deconvolution plugin was used to isolate the vector corresponding to the stain for fibrosis and then quantified using a threshold tool. For Histolab, areas of interest were selected by picking the color that corresponds to fibrosis, setting the threshold to account for slight variations in staining, and quantifying the selected area. Vvint, in all three methods, were averaged over all the images in each biopsy. We evaluated the level of agreement between methods using linear regression, correlation coefficients, and bias (computer methods - point counting). We also evaluated the relationship between measured Vvint, estimated glomerular filtration rate (eGFR), and Banff ci scores. Continuous data are reported as mean±standard deviation (SD).
Results: The average Banff ci score was 1.0±0.67 and average eGFR was 49.3±11.0mL/min/1.73m2. The average Vvint by the point counting method was 9.9±5.4%. Table 1 shows the correlation between the computer-based methods and point counting. There were no significant correlations between eGFR and measured fibrosis.
| Method | r | R² | Bias | SD |
| ImageJ | 0.68 | 0.47 | 10.1 | 8.1 |
| Histolab | 0.91 | 0.84 | 3.21 | 4.9 |