Abstract: SA-PO914
A Novel Morphometric Approach to Estimate Interstitial Fibrosis from Trichrome-Stained, Whole-Slide Images
Session Information
- Pathology and Lab Medicine - 2
October 26, 2024 | Location: Exhibit Hall, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Pathology and Lab Medicine
- 1800 Pathology and Lab Medicine
Authors
- Athavale, Ambarish, University of California San Diego, La Jolla, California, United States
- Patel, Tushar, University of Illinois Chicago, Chicago, Illinois, United States
- Kulkarni, Hemant, M&H Research LLC, San Antonio, Texas, United States
Background
Interstitial fibrosis (IF) is estimated by visual assessment which leads to significant interobserver variability. Using an image processing approach, we developed a novel algorithm (TRI_IF) to estimate IF from Trichrome whole slide images (WSI).
Methods
TRI_IF (developed in Python) masks out non-cortical area and estimates the proportion of blue stained area in the total cortical area (Figure 1A, 1B). 57 Trichrome WSI from the NEPTUNE study were used. Quality control excluded images with significant over/under staining and those beyond limits of agreement established by Bland-Altman plots. Cortex and glomeruli were manually annotated using QUPath. TRI_IF estimated IF was compared with NEPTUNE pathologists’ consensus estimate (NP_IF). Categorization of IF was done using the Banff classification for NP_IF and latent class thresholds (using item response theory) for TRI_IF. Comparisons were made using Pearson’s correlation for raw estimates, quadratically weighted Cohen’s kappa for clinical classification of IF and association with two clinical endpoints: a) composite of ESRD or 40% reduction of eGFR and b) slope of eGFR decline per year.
Results
Of 57 Trichrome WSI, 16 were excluded after quality control and 41 were included in the analysis. Pearson’s correlation between TRI_IF and NP_IF estimates was 0.69 (p<0.0001) and weighted Cohen’s kappa for categorized IF was 0.65 (p<0.0001, Figure 1C). Compared with low-to-mild IF (<35% for TRI_IF), Moderate-to-high IF ≥35% for TRI_IF) was associated with a higher risk of the composite outcome (Figure 1D) and faster decline of eGFR (Figure 1D, inset).
Conclusion
A novel algorithm (TRI_IF) for automated estimation of IF on Trichrome WSI showed good agreement with pathologist’s estimate of IF. Future studies need to validate this on a larger dataset.
Funding
- NIDDK Support