A large-scale retrospective study enabled deep-learning based pathological assessment of frozen procurement kidney biopsies to predict graft loss and guide organ utilization.

Lesion scores on procurement donor biopsies are commonly used to guide organ utilization for deceased-donor kidneys. However, frozen sections present challenges for histological scoring, leading to inter- and intra-observer variability and inappropriate discard. Therefore, we constructed deep-learning based models to recognize kidney tissue compartments in hematoxylin & eosin-stained sections from procurement needle biopsies performed nationwide in years 2011-2020. To do this, we extracted whole-slide abnormality features from 2431 kidneys and correlated with pathologists' scores and transplant outcomes. A Kidney Donor Quality Score (KDQS) was derived and used in combination with recipient demographic and peri-transplant characteristics to predict graft loss or assist organ utilization. The performance on wedge biopsies was additionally evaluated. Our model identified 96% and 91% of normal/sclerotic glomeruli respectively; 94% of arteries/arterial intimal fibrosis; 90% of tubules. Whole-slide features of Sclerotic Glomeruli (GS)%, Arterial Intimal Fibrosis (AIF)%, and Interstitial Space Abnormality (ISA)% demonstrated strong correlations with corresponding pathologists' scores of all 2431 kidneys, but had superior associations with post-transplant estimated glomerular filtration rates in 2033 and graft loss in 1560 kidneys. The combination of KDQS and other factors predicted one- and four-year graft loss in a discovery set of 520 kidneys and a validation set of 1040 kidneys. By using the composite KDQS of 398 discarded kidneys due to "biopsy findings", we suggest that if transplanted, 110 discarded kidneys could have had similar survival to that of other transplanted kidneys. Thus, our composite KDQS and survival prediction models may facilitate risk stratification and organ utilization while potentially reducing unnecessary organ discard.

Yi Z ，Xi C ，Menon MC ，Cravedi P ，Tedla F ，Soto A ，Sun Z ，Liu K ，Zhang J ，Wei C ，Chen M ，Wang W ，Veremis B ，Garcia-Barros M ，Kumar A ，Haakinson D ，Brody R ，Azeloglu EU ，Gallon L ，O'Connell P ，Naesens M ，Shapiro R ，Colvin RB ，Ward S ，Salem F ，Zhang W ... -  《-》

The role of procurement biopsies in acceptance decisions for kidneys retrieved for transplant.

Kasiske BL ，Stewart DE ，Bista BR ，Salkowski N ，Snyder JJ ，Israni AK ，Crary GS ，Rosendale JD ，Matas AJ ，Delmonico FL ... -  《-》

Reproducibility of Deceased Donor Kidney Procurement Biopsies.

Unfavorable histology on procurement biopsies is the most common reason for deceased donor kidney discard. We sought to assess the reproducibility of procurement biopsy findings. We compiled a continuous cohort of deceased donor kidneys transplanted at our institution from 1/1/2006 to 12/31/2016 that had at least one procurement biopsy performed, and excluded cases with missing biopsy reports and those used in multiorgan transplants. Suboptimal histology was defined as the presence of advanced sclerosis in greater than or equal to one biopsy compartment (glomeruli, tubules/interstitium, vessels). We calculated κ coefficients to assess agreement in optimal versus suboptimal classification between sequential biopsy reports for kidneys that underwent multiple procurement biopsies and used time-to-event analysis to evaluate the association between first versus second biopsies and patient and allograft survival. Of the 1011 kidneys included in our cohort, 606 (60%) had multiple procurement biopsies; 98% had first biopsy performed at another organ procurement organization and their second biopsy performed locally. Categorical agreement was highest for vascular disease (κ=0.17) followed by interstitial fibrosis and tubular atrophy (κ=0.12) and glomerulosclerosis (κ=0.12). Overall histologic agreement (optimal versus suboptimal) was κ=0.15. First biopsy histology had no association with allograft survival in unadjusted or adjusted analyses. However, second biopsy optimal histology was associated with a higher probability of death-censored allograft survival, even after adjusting for donor and recipient factors (adjusted hazard ratio, 0.50; 95% confidence interval, 0.34 to 0.75; P=0.001). Deceased donor kidneys that underwent multiple procurement biopsies often displayed substantial differences in histologic categorization in sequential biopsies, and there was no association between first biopsy findings and post-transplant outcomes.

Husain SA ，King KL ，Batal I ，Dube GK ，Hall IE ，Brennan C ，Stokes MB ，Crew RJ ，Carpenter D ，Alvarado Verduzco H ，Rosen R ，Coley S ，Campenot E ，Santoriello D ，Perotte A ，Natarajan K ，D'Agati VD ，Cohen DJ ，Ratner LE ，Markowitz G ，Mohan S ... -  《-》

Procurement Biopsies in the Evaluation of Deceased Donor Kidneys.

Biopsies taken at deceased donor kidney procurement continue to be cited as a leading reason for discard; however, the reproducibility and prognostic capability of these biopsies are controversial. We compiled a retrospective, single-institution, continuous cohort of deceased donor kidney transplants performed from 2006 to 2009. Procurement biopsy information-percentage of glomerulosclerosis, interstitial fibrosis/tubular atrophy, and vascular disease-was obtained from the national transplant database. Using univariable, multivariable, and time-to-event analyses for death-censored graft survival, we compared procurement frozen section biopsy reports with reperfusion paraffin-embedded biopsies read by trained kidney pathologists (n=270). We also examined agreement for sequential procurement biopsies performed on the same kidney (n=116 kidneys). For kidneys on which more than one procurement biopsy was performed (n=116), category agreement was found in only 64% of cases (κ=0.14). For all kidneys (n=270), correlation between procurement and reperfusion biopsies was poor: overall, biopsies were classified into the same category (optimal versus suboptimal) in only 64% of cases (κ=0.25). This discrepancy was most pronounced when categorizing percentage of glomerulosclerosis, which had 63% agreement (κ=0.15). Interstitial fibrosis/tubular atrophy and vascular disease had agreement rates of 82% (κ=0.13) and 80% (κ=0.15), respectively. Ninety-eight (36%) recipients died, and 56 (21%) allografts failed by the end of follow-up. Reperfusion biopsies were more prognostic than procurement biopsies (hazard ratio for graft failure, 2.02; 95% confidence interval, 1.09 to 3.74 versus hazard ratio for graft failure, 1.30; 95% confidence interval, 0.61 to 2.76), with procurement biopsies not significantly associated with graft failure. We found that procurement biopsies are poorly reproducible, do not correlate well with paraffin-embedded reperfusion biopsies, and are not significantly associated with transplant outcomes.

Carpenter D ，Husain SA ，Brennan C ，Batal I ，Hall IE ，Santoriello D ，Rosen R ，Crew RJ ，Campenot E ，Dube GK ，Radhakrishnan J ，Stokes MB ，Sandoval PR ，D'Agati V ，Cohen DJ ，Ratner LE ，Markowitz G ，Mohan S ... -  《-》

Deep learning identified pathological abnormalities predictive of graft loss in kidney transplant biopsies.

Interstitial fibrosis, tubular atrophy, and inflammation are major contributors to kidney allograft failure. Here we sought an objective, quantitative pathological assessment of these lesions to improve predictive utility and constructed a deep-learning-based pipeline recognizing normal vs. abnormal kidney tissue compartments and mononuclear leukocyte infiltrates. Periodic acid- Schiff stained slides of transplant biopsies (60 training and 33 testing) were used to quantify pathological lesions specific for interstitium, tubules and mononuclear leukocyte infiltration. The pipeline was applied to the whole slide images from 789 transplant biopsies (478 baseline [pre-implantation] and 311 post-transplant 12-month protocol biopsies) in two independent cohorts (GoCAR: 404 patients, AUSCAD: 212 patients) of transplant recipients to correlate composite lesion features with graft loss. Our model accurately recognized kidney tissue compartments and mononuclear leukocytes. The digital features significantly correlated with revised Banff 2007 scores but were more sensitive to subtle pathological changes below the thresholds in the Banff scores. The Interstitial and Tubular Abnormality Score (ITAS) in baseline samples was highly predictive of one-year graft loss, while a Composite Damage Score in 12-month post-transplant protocol biopsies predicted later graft loss. ITASs and Composite Damage Scores outperformed Banff scores or clinical predictors with superior graft loss prediction accuracy. High/intermediate risk groups stratified by ITASs or Composite Damage Scores also demonstrated significantly higher incidence of estimated glomerular filtration rate decline and subsequent graft damage. Thus, our deep-learning approach accurately detected and quantified pathological lesions from baseline or post-transplant biopsies and demonstrated superior ability for prediction of post-transplant graft loss with potential application as a prevention, risk stratification or monitoring tool.

Yi Z ，Salem F ，Menon MC ，Keung K ，Xi C ，Hultin S ，Haroon Al Rasheed MR ，Li L ，Su F ，Sun Z ，Wei C ，Huang W ，Fredericks S ，Lin Q ，Banu K ，Wong G ，Rogers NM ，Farouk S ，Cravedi P ，Shingde M ，Smith RN ，Rosales IA ，O'Connell PJ ，Colvin RB ，Murphy B ，Zhang W ... -  《-》