Utility of Whole Slide Imaging Algorithms in the Assessment of Renal Allograft Biopsies: Does Automated Analysis Provide a Better Assessment of Donor Specific Antibody Than Human Scoring?
Lauren N Stuart, Gizem Tumer, Tiffany Roberts-Wilson, Howard M Gebel, Robert A Bray, Alton B Farris. Emory University, Atlanta, GA
Background: Antibody mediated rejection (AMR) in renal transplants is often attributed to donor specific antibodies (DSAs). Circulating DSAs lead to complement split product C4d deposition along peritubular capillary endothelium, as highlighted by C4d immunohistochemistry (IHC). The extent of C4d deposition has been scored according to the Banff criteria (C4d1-minimal,C4d2-focal,C4d3-diffuse). However, the relationship between C4d staining extent and circulating DSA quantity has not been fully characterized; thus, we sought to determine the correlations amongst human scoring, whole slide image (WSI) analysis, and DSA-specific mean fluorescence intensity (MFI) levels.
Design: Renal allograft biopsies with a confirmed DSA were selected from our files. The initial C4d scoring was verified on repeat pathologic review of the C4d IHC stained slides. These cases were scanned using an Aperio WSI scanner and analyzed with the Aperio microvessel density algorithm, which measured 18 vascular parameters per case. HLA class I and II DSAs for each biopsy were identified and a cumulative, DSA-specific MFI was calculated.
Results: C4d scoring and WSI analysis were compared to the cumulative MFI value (range: 2,600–385,000) for each case. The cumulative MFI was found to be moderately correlated with human C4d scoring (r=0.557,p=0.0071). The cumulative MFI only weakly correlated with automated WSI analysis, with median vessel perimeter being the strongest correlating parameter (r=0.367,p=0.0935). In order to achieve a more accurate reflection of the MFI data, the cases were then stratified into 4 ranks: “non-contributory”, “low”, “intermediate”, and “high.” This ranked MFI was also found to be moderately correlated with human C4d scoring (r =0.643,p=0.0011) and showed improved correlation with automated WSI analysis. Amongst the WSI analysis output parameters, median vessel perimeter and median vascular area were most correlated with ranked MFI (r=0.547,p=0.0084 and r=0.523,p=0.0125, respectively). Weaker correlations were seen with median lumen area and mean vessel wall thickness (r=0.505,p=0.0165 and r=0.386,p=0.0759, respectively).
Conclusions: Human scoring of C4d deposition provided a better correlation with DSA MFI than automated image analysis. However, more sophisticated WSI equipment and algorithms may lead to more accurate assessments in the future. Ultimately, improved histology examination and DSA detection could lead to enhanced recognition and therapy of AMR.
Category: Kidney (does not include tumors)
Monday, March 4, 2013 9:30 AM
Poster Session I Stowell-Orbison/Surgical Pathology/Autopsy Awards Poster Session # 254, Monday Morning