1
|
Glass M, Ji Z, Davis R, Pavlisko EN, DiBernardo L, Carney J, Fishbein G, Luthringer D, Miller D, Mitchell R, Larsen B, Butt Y, Bois M, Maleszewski J, Halushka M, Seidman M, Lin CY, Buja M, Stone J, Dov D, Carin L, Glass C. A machine learning algorithm improves the diagnostic accuracy of the histologic component of antibody mediated rejection (AMR-H) in cardiac transplant endomyocardial biopsies. Cardiovasc Pathol 2024; 72:107646. [PMID: 38677634 DOI: 10.1016/j.carpath.2024.107646] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 04/03/2023] [Revised: 04/15/2024] [Accepted: 04/16/2024] [Indexed: 04/29/2024] Open
Abstract
BACKGROUND Pathologic antibody mediated rejection (pAMR) remains a major driver of graft failure in cardiac transplant patients. The endomyocardial biopsy remains the primary diagnostic tool but presents with challenges, particularly in distinguishing the histologic component (pAMR-H) defined by 1) intravascular macrophage accumulation in capillaries and 2) activated endothelial cells that expand the cytoplasm to narrow or occlude the vascular lumen. Frequently, pAMR-H is difficult to distinguish from acute cellular rejection (ACR) and healing injury. With the advent of digital slide scanning and advances in machine deep learning, artificial intelligence technology is widely under investigation in the areas of oncologic pathology, but in its infancy in transplant pathology. For the first time, we determined if a machine learning algorithm could distinguish pAMR-H from normal myocardium, healing injury and ACR. MATERIALS AND METHODS A total of 4,212 annotations (1,053 regions of normal, 1,053 pAMR-H, 1,053 healing injury and 1,053 ACR) were completed from 300 hematoxylin and eosin slides scanned using a Leica Aperio GT450 digital whole slide scanner at 40X magnification. All regions of pAMR-H were annotated from patients confirmed with a previous diagnosis of pAMR2 (>50% positive C4d immunofluorescence and/or >10% CD68 positive intravascular macrophages). Annotations were imported into a Python 3.7 development environment using the OpenSlide™ package and a convolutional neural network approach utilizing transfer learning was performed. RESULTS The machine learning algorithm showed 98% overall validation accuracy and pAMR-H was correctly distinguished from specific categories with the following accuracies: normal myocardium (99.2%), healing injury (99.5%) and ACR (99.5%). CONCLUSION Our novel deep learning algorithm can reach acceptable, and possibly surpass, performance of current diagnostic standards of identifying pAMR-H. Such a tool may serve as an adjunct diagnostic aid for improving the pathologist's accuracy and reproducibility, especially in difficult cases with high inter-observer variability. This is one of the first studies that provides evidence that an artificial intelligence machine learning algorithm can be trained and validated to diagnose pAMR-H in cardiac transplant patients. Ongoing studies include multi-institutional verification testing to ensure generalizability.
Collapse
Affiliation(s)
- Matthew Glass
- Duke Division of Artificial Intelligence and Computational Pathology, Duke University Medical Center, Durham NC, USA; Department of Anesthesiology, Duke University Medical Center, Durham NC, USA
| | - Zhicheng Ji
- Department of Biostatistics and Bioinformatics, Duke School of Medicine, Durham NC, USA
| | - Richard Davis
- Department of Pathology, Duke University Medical Center, Durham NC, USA
| | - Elizabeth N Pavlisko
- Duke Division of Artificial Intelligence and Computational Pathology, Duke University Medical Center, Durham NC, USA; Department of Pathology, Duke University Medical Center, Durham NC, USA
| | - Louis DiBernardo
- Department of Pathology, Duke University Medical Center, Durham NC, USA
| | - John Carney
- Department of Pathology, Duke University Medical Center, Durham NC, USA
| | - Gregory Fishbein
- Department of Pathology, University of California at Los Angeles, Los Angeles CA, USA
| | - Daniel Luthringer
- Department of Pathology and Laboratory Medicine, Cedars-Sinai Medical Center, Los Angeles CA, USA
| | - Dylan Miller
- Department of Pathology, Intermountain Healthcare, Salt Lake City UT, USA
| | - Richard Mitchell
- Department of Pathology, Brigham and Women's Hospital, Boston MA, USA
| | - Brandon Larsen
- Department of Pathology and Laboratory Medicine, Mayo Clinic, Phoenix AZ, USA
| | - Yasmeen Butt
- Department of Pathology and Laboratory Medicine, Mayo Clinic, Phoenix AZ, USA
| | - Melanie Bois
- Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester MN, USA
| | - Joseph Maleszewski
- Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester MN, USA
| | - Marc Halushka
- Department of Pathology, Johns Hopkins University School of Medicine, Baltimore MD, USA
| | - Michael Seidman
- Department of Pathology, University Health Network, Toronto ON, CA
| | - Chieh-Yu Lin
- Department of Pathology and Immunology, Washington University, St. Louis MO, USA
| | - Maximilian Buja
- Department of Pathology and Laboratory Medicine, The University of Texas Health Science Center at Houston, Houston TX, USA
| | - James Stone
- Department of Pathology, Massachusetts General Hospital, Boston MA, USA
| | - David Dov
- Duke Division of Artificial Intelligence and Computational Pathology, Duke University Medical Center, Durham NC, USA; Pratt School of Engineering, Department of Electrical and Computer Engineering, Duke University, Durham NC, USA
| | - Lawrence Carin
- Duke Division of Artificial Intelligence and Computational Pathology, Duke University Medical Center, Durham NC, USA; Pratt School of Engineering, Department of Electrical and Computer Engineering, Duke University, Durham NC, USA
| | - Carolyn Glass
- Duke Division of Artificial Intelligence and Computational Pathology, Duke University Medical Center, Durham NC, USA; Department of Pathology, Duke University Medical Center, Durham NC, USA.
| |
Collapse
|
2
|
Kim PJ, Cusi V, Cardenas A, Tada Y, Vaida F, Wettersten N, Chak J, Bijlani P, Pretorius V, Urey MA, Morris GP, Lin G. Antibody Mediated Rejection is not Associated with Worse Survival in Adherent Heart Transplant Patients in the Contemporary Era. MEDRXIV : THE PREPRINT SERVER FOR HEALTH SCIENCES 2023:2023.12.01.23299311. [PMID: 38106112 PMCID: PMC10723500 DOI: 10.1101/2023.12.01.23299311] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Grants] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 12/19/2023]
Abstract
Background C4d immunostaining of surveillance endomyocardial biopsies (EMB) and testing for donor specific antibodies (DSA) are routinely performed in the first year of heart transplantation (HTx) in adult patients. C4d and DSA positivity have not been evaluated together with respect to clinical outcomes in the contemporary era (2010-current). Methods This was a single center, retrospective study of consecutive EMBs performed between November 2010 and April 2023. The primary objective was to determine whether history of C4d and/or DSA positivity could predict death, cardiac death, or retransplant. Secondary analyses included cardiac allograft dysfunction and cardiac allograft vasculopathy. Cox proportional hazards models were used for single predictor and multipredictor analyses. Results A total of 6,033 EMBs from 519 HTx patients were reviewed for the study. There was no significant difference (p = 0.110) in all-cause mortality or cardiac retransplant between four groups: C4d+/DSA+, C4d+/DSA-, C4d-/DSA+, and C4d-/DSA-. The risk for cardiac mortality or retransplant was significantly higher in C4d+/DSA+ versus C4d-/DSA- patients (HR = 4.73; pc = 0.042) but not significantly different in C4d+/DSA- versus C4d-/DSA- patients (pc = 1.000). Similarly, the risk for cardiac allograft dysfunction was significantly higher in C4d+/DSA+ versus C4d-/DSA- patients (HR 3.26; pc = 0.001) but not significantly different in C4d+/DSA- versus C4d-/DSA- patients (pc = 1.000). Accounting for nonadherence, C4d/DSA status continued to predict cardiac allograft dysfunction but no longer predicted cardiac death or retransplant. Conclusions Medically adherent C4d+/DSA+ HTx patients show significantly greater risk for cardiac allograft dysfunction but not cardiac mortality or retransplant. In contrast, C4d+/DSA- patients represent a new immunopathologic group with a clinical course similar to that of HTx patients without antibody mediated rejection.
Collapse
Affiliation(s)
| | | | - Ashley Cardenas
- Department of Pathology, University of California, San Diego, California, USA
| | | | - Florin Vaida
- Department of Family Medicine and Public Health, UC San Diego, La Jolla, CA
| | - Nicholas Wettersten
- Cardiology Section, Veterans Affairs San Diego Healthcare System, San Diego, CA
| | | | | | - Victor Pretorius
- Division of Cardiovascular and Thoracic Surgery, Department of Surgery, University of California, San Diego, California, USA
| | | | - Gerald P Morris
- Department of Pathology, University of California, San Diego, California, USA
| | - Grace Lin
- Department of Pathology, University of California, San Diego, California, USA
| |
Collapse
|
3
|
Skougaard M, Bærentzen S, Eiskjær H, Koefoed-Nielsen P. Eosinophilic infiltration as the initial trace of acute mixed cellular and antibody mediated rejection in a heart transplant patient with concomitant immense epitope-associated HLA-antibody production: a case report. Front Immunol 2023; 14:1207373. [PMID: 37744343 PMCID: PMC10516220 DOI: 10.3389/fimmu.2023.1207373] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/17/2023] [Accepted: 07/18/2023] [Indexed: 09/26/2023] Open
Abstract
Acute mixed cellular and antibody-mediated rejection (MR) has an estimated prevalence of 7.8%. However, knowledge of MR immune pathogenesis in cardiac graft rejection remains sparse. We report a case of acute MR in a heart transplant patient with a mutation in the MYH7 gene encoding the protein β-myosin heavy chain, resulting in familial hypertrophic cardiomyopathy. The patient presented with substantial eosinophilic infiltration and extensive production of Human Leukocyte Antigen (HLA)-antibodies associated with shared epitopes. Eosinophilic infiltration in the endo- and myocardium was diagnosed in routine post-transplant biopsies stained with hematoxylin-eosin on day 6 after transplantation. On day 27, the patient presented with dyspnea, weight gain, increased pro-brain natriuretic peptide, and was hospitalized due to suspected acute rejection. Endomyocardial biopsies showed eosinophils in endo- and myocardium with additional lymphocytes and hyperplastic endothelium. Immunohistochemistry, including CD31/CD68 double stain confirmed endothelium-associated macrophages in capillaries and severe C4d positivity in the capillaries and endocardial endothelium. Lymphocytes were identified as primarily CD45+/CD3+ T cells with a concomitant few CD45+/CD20+ B cells. HLA-antibody analysis demonstrated a significant increase in 13 HLA-antibodies present in pre-transplant-serum, of which anti-B7 was donor-specific, and 23 strong de-novo HLA-class I antibodies of which anti-B62 was donor-specific. 72% of HLA-antibodies, including the two donor-specific antibodies, shared the same HLA antigen epitope; 43P+69A or 163L+167W. This is a case reporting both HLA-antibody and pathohistological data indicating the need for better understanding of interactions between cellular and antibody-mediated immune response mechanisms in graft rejection, and the significance of pre-transplant donor-specific antibodies during immunological pre-transplant risk assessment.
Collapse
Affiliation(s)
- Marie Skougaard
- Department of Clinical Immunology, Aarhus University Hospital, Aarhus, Denmark
| | - Steen Bærentzen
- Department of Pathology, Aarhus University Hospital, Aarhus, Denmark
| | - Hans Eiskjær
- Department of Cardiology, Aarhus University Hospital, Aarhus, Denmark
| | | |
Collapse
|
4
|
Cai Q, Moore SA, Hendricks AR, Torrealba JR. Upregulation of Endothelial HLA Class II is a Marker of Antibody-Mediated Rejection in Heart Allograft Biopsies. Transplant Proc 2020; 52:1192-1197. [PMID: 32197864 DOI: 10.1016/j.transproceed.2020.01.049] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/30/2019] [Accepted: 01/22/2020] [Indexed: 12/24/2022]
Abstract
In 2013, the International Society of Heart and Lung Transplant (ISHLT) introduced the working classification for pathologic changes associated with antibody-mediated rejection (AMR) of the heart allograft, known as pathologic AMR (pAMR). With 2 components associated with AMR, histopathologic changes) and immunopathologic markers, the proposed classification also suggests the use of class II HLA as a marker of endothelial integrity. It is known that during allograft rejection, endothelial cells are activated, therefore, we hypothesized that endothelial class II HLA rather than a marker of mere endothelial presence, is a marker of endothelial activation and becomes upregulated in AMR. Eight hundred thirty-eight heart allograft biopsies, collected from January 2016 to September 2018 at a single institution from patients with a current or recent diagnosis of AMR, were evaluated for both histopathologic and immunopathologic changes of AMR. Biopsies were labeled with immunofluorescence with antibodies against C4d and for immunohistochemistry with antibodies against C3d, CD68, and class II HLA. ISHLT criteria were used to classify the biopsies, and for class II HLA, both the percentage and the stain intensity were evaluated. Biopsies (74.8%) from our cohort showed either histopathologic pAMR-1, immunopathologic pAMR-1, or combined histopathologic and immunopathologic pAMR-2 evidence of AMR. Expression of endothelial HLA class II was significantly correlated with the diagnosis of AMR by percentage area (P < .0001) and intensity of staining (P < .0001). The diagnosis of AMR significantly correlated with moderate (+2) and strong (+3) staining intensity for class II HLA as follows: histopathologic and immunopathologic pAMR-2 with odds ratio (OR) = 28.3 (P < .0001);histopathologic pAMR-1 alone with OR = 22.7 (P < .0001); and immunopathologic pAMR-1 alone with OR = 32.6 (P < .0001). Interestingly, our study also suggested that the inclusion of C4d focally positive cases also significantly correlates with the diagnosis of AMR (P < .003). We confirmed our hypothesis that in heart allograft biopsies, there is a spectrum of both percentage and intensity of HLA class II expression due to endothelial activation, and that class II HLA by immunohistochemistry is a marker significantly correlated with the diagnosis of AMR. In addition, the group of focally positive C4d biopsies (10%-50%) should be considered positive for the immunopathologic component of the 2013 ISHLT classification, as this group of biopsies also correlated with the diagnosis of AMR.
Collapse
Affiliation(s)
- Qi Cai
- University of Texas Southwestern Medical Center, Dallas, Texas
| | | | | | | |
Collapse
|