Browsing by Subject "digital pathology algorithm"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Open Access
    Renal allograft biopsies at Groote Schuur Hospital: a histopathologic descriptive study with molecular insights
    (2025) Lunn, Jarryd; Price, Brendon; Chetty, Dharshnee; Ikumi, Nadia
    Background: Kidney transplantation is the definitive treatment for end-stage kidney disease. Immune-mediated rejection remains a barrier to success. It is diagnosed through the Banff classification, which incorporates histopathology and biomarkers (C4d/donor specific antibodies (DSA)). In resource-limited settings, DSA testing can be challenging, necessitating reliable alternatives. Aim: This study evaluated: (1) rejection patterns at our hospital; (2) the impact of the Banff 2022 criteria with a computer-assisted tool; and (3) the utility of C4d as a predictor of DSA status. Methods: We analysed 197 for-cause historic biopsy reports between 2015-2022 for details of rejection- and non-rejection pathologies, Banff lesion scores and DSA status. A computer- based tool was used on historic data to re-calculate Banff 2022 classification diagnoses, which were compared to historic diagnoses. Logistic regression assessed C4d as a predictor of DSA. Results: The cohort showed a male predominance (59.3%). Sixty-three percent of cases showed non-rejection pathology, with acute tubular injury and pyelonephritis being the most frequent. TCMR was the most common form of rejection (17.3%), with AMR being the least common (7.6%). The computer-based tool demonstrated agreement of 92.4% for AMR/TCMR and 84.6% of borderline TCMR, but was confounded by non-rejection pathologies. C4d predicted DSA-positivity with 95% specificity but only 29.5% sensitivity. Conclusion: The Banff 2022 criteria were additive in rejection diagnosis, with a computer-based tool acting as a guide but not a pure diagnostic tool. The high specificity of C4d makes it valuable where DSA testing is limited. Contribution: This study validates the role of the Banff 2022 in our setting, aided by a computer-based tool that aims to decrease logical- and transcription errors when using the complex Banff classification. It also demonstrates C4d's role as a practical DSA proxy, offering actionable solutions in resource-limited settings.