Abstract: SA-PO002
Computational Assessment of Glomerular Basement Membrane Width and Podocyte Foot Process Width in an Animal Model of Podocytopathy
Session Information
- Bioengineering
November 05, 2022 | Location: Exhibit Hall, Orange County Convention Center‚ West Building
Abstract Time: 10:00 AM - 12:00 PM
Category: Bioengineering
- 300 Bioengineering
Authors
- Laudon, Aksel David, Boston University College of Engineering, Boston, Massachusetts, United States
- Kim, Connor, Boston University College of Engineering, Boston, Massachusetts, United States
- Qian, Yingzhe, Boston University College of Engineering, Boston, Massachusetts, United States
- Wang, Zhaoze, Boston University College of Engineering, Boston, Massachusetts, United States
- Ye, Qin, Boston University College of Engineering, Boston, Massachusetts, United States
- Kolachalama, Vijaya B., Boston University School of Medicine, Boston, Massachusetts, United States
- Henderson, Joel M., Boston University School of Medicine, Boston, Massachusetts, United States
- Lu, Weining, Boston University School of Medicine, Boston, Massachusetts, United States
Background
Transmission electron microscopy (TEM) measurements of glomerular basement membrane (GBM) width and podocyte foot process width (FPW) are important diagnostic tools for podocytopathy. Currently, these measurements in TEM images are performed manually, which limits experimental ultrastructural analysis. Here, we describe a computational approach to measure GBM width and podocyte FPW of healthy and pathological kidney specimens using TEM images from Integrin-Linked Kinase podocyte-specific knockout mice (ILK cKO), an animal model of podocytopathy.
Methods
We obtained TEM images from five wildtype (WT) littermates and five ILK cKO mice at 4 weeks of age. We developed a computational approach with two stages: a pre-trained U-Net-based machine learning framework for GBM segmentation and an image processing algorithm for GBM width and FPW measurement. We evaluated its performance in a five-fold cross-validation study. Segmentation accuracy was assessed in terms of Jaccard index, the ratio of the size of intersection to the size of union for an image’s predicted GBM and manually annotated reference GBM label. Automated measurements were compared to corresponding manual measurements between WT and ILK cKO mice.
Results
The cross-validation study resulting mean TEM image Jaccard index was 0.56 for WT and 0.66 for ILK cKO. Automated mean GBM width and FPW closely matched manual measurements for WT (p = 0.63, p = 0.31) but differed for ILK cKO specimens (p = 0.06, p = 0.06). Automated GBM width (p = 0.008) and FPW (p = 0.03) measurements were significantly wider for ILK cKO than WT mice, which aligns with known morphology.
Conclusion
We developed a machine-learning based approach to measure GBM width and FPW in TEM images for an animal model of podocytopathy. Our results suggest that automated measurements could distinguish healthy from pathological kidneys. Our tool provides high-throughput, objective morphological analysis and could potentially facilitate podocytopathy and other glomerular disease research and diagnosis in the future.