Abstract: FR-OR56
Quantitative Characterization of Nonsclerotic Glomeruli Is Prognostic of Clinical Outcomes in Proteinuric Diseases
Session Information
- Kidney Pathology: From Classic Clinicopathologic Studies to Computational Pathology
November 03, 2023 | Location: Room 109, Pennsylvania Convention Center
Abstract Time: 05:15 PM - 05:24 PM
Category: Pathology and Lab Medicine
- 1800 Pathology and Lab Medicine
Authors
- Ambekar, Akhil, Department of Pathology, Division of AI & Computational Pathology, Duke University, Durham, North Carolina, United States
- Liu, Qian, Children’s Hospital of Philadelphia Research Institute, Philadelphia, Pennsylvania, United States
- Zee, Jarcy, Children’s Hospital of Philadelphia Research Institute, Philadelphia, Pennsylvania, United States
- Wang, Bangchen, Department of Pathology, Division of AI & Computational Pathology, Duke University, Durham, North Carolina, United States
- Cassol, Clarissa Araujo, Arkana Laboratories, Little Rock, Arkansas, United States
- Mariani, Laura H., Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan, United States
- Holzman, Lawrence B., Department of Medicine, Division of Nephrology and Hypertension, University of Pennsylvania, Philadelphia, Pennsylvania, United States
- Hodgin, Jeffrey B., Department of Pathology, University of Michigan, Ann Arbor, Michigan, United States
- Madabhushi, Anant, Department of Biomedical Engineering, Emory University and Georgia Institute of Technology, Atlanta, Georgia, United States
- Lafata, Kyle Jon, Department of Radiation Oncology, Duke University, Durham, North Carolina, United States
- Barisoni, Laura, Department of Pathology, Division of AI & Computational Pathology, Duke University, Durham, North Carolina, United States
- Janowczyk, Andrew, Department of Biomedical Engineering, Emory University and Georgia Institute of Technology, Atlanta, United States
Background
Visual assessment of segmentally (SS), globally (GS) and non-GS/SS glomeruli is used clinically for the diagnosis and prognostication of glomerular diseases. We hypothesized that encoded in non-GS/SS glomeruli are computationally derivable subvisual features prognostic of outcome.
Methods
N=186 (99 MCD/MCD-like, 87 FSGS) patients from the NEPTUNE/CureGN cohorts with at least one PAS whole slide image (WSI) containing ≥4 non-GS/SS glomeruli were included. A previously validated pipeline for glomerular segmentation and classification was applied to WSIs, yielding 272 GS, 113 SS, and 2661 non-GS/SS glomeruli. Percent of SS and GS was calculated. From non-GS/SS glomeruli, 108 intensity, shape, and texture features were computed. Patient-level summary statistics were produced using mean, standard deviation, kurtosis, minimum, maximum, and median. Maximum Relevance Minimum Redundancy (MRMR) selected the 10 most prognostic features for time from biopsy to disease progression (≥40% eGFR decline with last eGFR<90 or kidney failure), and to first complete proteinuria remission (UPCR<0.3). Ridge regression models estimated prognostic accuracy of non-GS/SS glomerular features.
Results
Prognostic features (Table 1) reflect the heterogeneity of intra-glomerular organization. These features were the most prognostic of disease progression compared to other models and increased the prognostic accuracy of both clinical outcomes when added to conventional parameters (Table 2).
Conclusion
Computational methods allow for extraction from non-GS/SS glomeruli of information prognostic of clinical outcomes above and beyond conventional methods. This novel approach may enable early diagnosis and facilitate risk-stratification independently from the presence of conventional diagnostic lesions of sclerosis.
Funding
- NIDDK Support