ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on X

Kidney Week

Abstract: SA-PO0002

Panoptic Segmentation-Based Method for Automated Analysis of Kidney Transmission Electron Microscopy (TEM) Images Across Species and Disease Models

Session Information

Category: Artificial Intelligence, Digital Health, and Data Science

  • 300 Artificial Intelligence, Digital Health, and Data Science

Authors

  • Zou, Anqi, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Tan, Winston, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Ji, Jiayi, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Fan, Xueping, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Dodd, Laura, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Oei, Emily, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Chen, Hui, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Liu, Yu-Chen, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Rojas-Miguez, Florencia, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Henderson, Joel M., Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Zhang, Chao, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
  • Lu, Weining, Boston University Chobanian & Avedisian School of Medicine, Boston, Massachusetts, United States
Background

Abnormalities in the glomerular basement membrane (GBM) and podocyte foot process (PFP) widths are key diagnostic indicators in proteinuric kidney diseases. Time-consuming manual measurement is routinely used in pathological diagnosis and animal model research. Due to a lack of high-quality labeled data and properly designed models, there are currently no reliable tools available to automatically and accurately measure the GBM and PFP widths.

Methods

To overcome the challenge, we manually labelled GBM and PFP in over 1,000 TEM images from human, mouse and rat samples to create a high-quality training dataset. We then developed TEAMKidney, a deep learning-based framework that utilizes the Mask2Former with an HRNet backbone, selected for its ability to process high-resolution TEM images. To improve generalization and robustness, the model uses self-training with pseudo-label refinement across species.

Results

We evaluated TEAMKidney on TEM images from diverse human diseases, including Fabry and diabetic kidney disease, ILK and COL4A3 knockout mouse models, and the rat Passive Heymann Nephritis model. Our method outperforms all existing approaches in GBM and PFP width quantification, supported by strong segmentation performance, with accuracy of 86.154 (segmentation quality) for GBM and 83.225 (panoptic quality) for PFP. Additionally, our method produces highly consistent results across varying resolutions, magnifications, and instruments. Overall, TEAMKidney demonstrates a strong agreement with diagnostic measurements made by experienced pathologists.

Conclusion

TEAMKidney can precisely segment ultra-structures from glomerular TEM images and accurately quantify the GBM and PFP widths. This is the first deep learning framework for TEM images applicable across species and diseases, holding promises for facilitating clinical-grade quantitative pathology in kidney disease diagnosis.

Figure1. Workflow of the proposed TEAMKidney framework.

Funding

  • NIDDK Support

Digital Object Identifier (DOI)