Abstract: TH-PO628
Prediction of Response to Intensified Immunosuppression in Childhood Steroid-Resistant Nephrotic Syndrome by Multimodal Machine Learning
Session Information
- Membranous Nephropathy, FSGS, and Minimal Change Disease
October 24, 2024 | Location: Exhibit Hall, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Glomerular Diseases
- 1402 Glomerular Diseases: Clinical, Outcomes, and Therapeutics
Authors
- Schaefer, Franz, Universitat Heidelberg, Heidelberg, Baden-Württemberg, Germany
- Morello, William, Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico, Milan, Lombardia, Italy
- Vo, Huy Quoc, University of Houston, Houston, Texas, United States
- Pedraza, Anibal, Universidad de Castilla-La Mancha, Ciudad Real, Spain
- Altini, Nicola, Politecnico di Bari, Bari, Italy
- Saleem, Moin A., University Hospitals Bristol and Weston NHS Foundation Trust, Bristol, United Kingdom
- Mohan, Chandra, University of Houston, Houston, Texas, United States
- Becker, Jan U., Universitat zu Koln, Cologne, Nordrhein-Westfalen, Germany
Background
There are no established models to predict the varied response to intensified immunosuppression in steroid-resistant nephrotic syndrome (SRNS). Multimodal machine learning, integrating clinical and genetic data with nephropathology imaging holds great promise to deliver such classifiers with perfect reproducibility. Here, we present such a theranostic classifier based on a multi-centric dataset obtained from the PodoNet registry cohort.
Methods
Data from 201 SRNS patients with available clinical data and whole slide images of kidney biopsies were collected from 14 European pediatric nephrology centres. The clinical input data contained 19 parameters including genetic mutation (yes/no) and eGFR at biopsy; the ground truth theranostic endpoint was treatment response as no (n=114), partial (n=42) and complete remission (n=45) within 6 months as defined by serum albumin and proteinuria changes. We trained our proprietary multimodal MorphSet++ architecture in a weakly supervised fashion. MorphSet++ integrates the clinical data vector from a shallow network at various stages with the transformer-based deep network analysis of the histopathology imaging data. Results are given as means after 5-fold internal cross-validation.
Results
Mean AUC was 80% for complete, 77% for partial and and 81% for no remission. The mean positive predictive value was 89.7, 89.8, 89.3 % and the mean negative predictive value 94.8, 93.4 and 97.6%. Mean sensitivity was 84.2, 67.1, and 98.5% and mean specificity 96.8, 98.4, and 83.9%. Mean balanced accuracy was 90.5%, 82.8%, and 91.2% for no, partial and complete remission, respectively. Adding the results of image analysis improved model performance as compared to clinical/genetic data input alone.t of SRNS.
Conclusion
Our MorphSet++ architecture shows promising results as a theranostic tool to predict response to immunosuppression in SRNS. MorphSet++ allows for rapid up-scaling with additional, larger datasets for even better and more robust performance. This might lead to a re-appraisal of nephropathology in the clinical management of SRNS.
Funding
- Government Support – Non-U.S.