Kidney Week 2025 Annual Meeting-Artificial Intelligence and Data Science Transforming Kidney Care: From Algorithms to Action

Video Transcription

Video Summary

This session highlighted how artificial intelligence and data science are moving “from algorithms to actions” in kidney care, spanning ICU prediction, dialysis behavior phenotyping, CKD forecasting, transplant decision support, multi-omics target discovery, spatial pathology segmentation, proteomic disease classification, nephron-scale mapping, and organ-on-chip modeling.<br /><br />Dr. Joanna Zhang presented a temporal XGBoost model using CRRTNet registry data to predict 7‑day renal replacement therapy (RRT)-free survival in ICU patients on CRRT. By combining static variables with recent EHR trends (via TSFRESH rolling windows), the best deployable approach used a 1‑day window (AUROC ~0.89, AUPRC ~0.83). SHAP interpretability highlighted hemodynamics, vasopressor dose, fluid balance, and urine output as key drivers.<br /><br />Dr. Hikaru Kadono used accelerometer data and UMAP+k-means clustering in 1,030 hemodialysis patients to derive “digital physical activity phenotypes.” An “MVPA-mixed” phenotype (despite similar total activity) had ~30% lower mortality than an “LPA-dominant” phenotype.<br /><br />Dr. David Shang validated and optimized a GRU-based CKD progression model (KDPP) trained in Taiwan, showing performance drops on U.S. Mayo data but recovery with fine-tuning; importantly, it remained usable even when UACR was missing, outperforming KFRE.<br /><br />Dr. Luis Rodriguez described ComPreps, integrating AI digital pathology (computational Remuzzi scoring + radiomics/pathomics) with donor clinical data to predict 12‑month eGFR and delayed graft function, potentially reducing unnecessary kidney discard.<br /><br />Subsequent talks covered: multimodal kidney atlases and knowledge graphs for drug target nomination (including LLM-assisted prioritization), deep-learning segmentation of tubules/immune infiltrates in lupus nephritis spatial proteomics, plasma proteomics plus elastic-net models to classify primary glomerular diseases (weaker for heterogeneous FSGS), 3D nephron tracking from cleared tissue light-sheet imaging, and an iPSC-derived “glomerulus-on-a-chip” for mechanobiology and disease modeling (e.g., Alport syndrome) aimed at drug testing and reducing animal use.

Asset Subtitle

Moderator(s): Karin Bergling, Wisit Cheungpasitporn

Presentation(s):

Temporal Machine-Learning Model for Predicting Seven-Day RRT-Free Survival in Critically Ill Patients on Continuous Renal Replacement Therapy - Joanna Yang
Machine Learning-Driven Digital Phenotyping of Accelerometer-Based Physical Activity Patterns and Mortality in Patients on Hemodialysis: The PROMOTE Study - Hikaru Kadono
Real-World Validation and Optimization of CKD Progression Prediction Models Using US Mayo Clinic Data - David Chang
Integrating Pathomic and Clinical Data Using Artificial Intelligence (AI) Models from Kidney Deceased Donors to Reduce Organ Discard and Improve Transplant Outcomes - Luis Rodrigues
NephroAIX: Large Language Model (LLM)-Guided Knowledge Graph for Explainable Target Discovery in Human Kidney Diseases Using Single-Cell Data - Sikander Hayat
TRIMS: Deep-Learning Segmentation of Renal Tubules and Immune Cells in 32-Marker Spatial Proteomic Images from Lupus Nephritis Biopsies - Mahith Gottipati
Large-Scale Systemic Proteome Profiling of More Than 5,000 Proteins to Differentiate Primary Glomerular Diseases - Jae-ik Oh
High-Throughput Nephron Tracking Reveals Hierarchical Organization in Mouse Kidneys - Joshua Vaughan
Human Induced Pluripotent Stem Cell (iPSC)-Derived Glomerulus-on-a-Chip for Disease Modeling and Drug Testing - Laura Locatelli

Note: Continuing education credits are not being offered for this session.

Meta Tag

Date 11/7/2025

Session ID 519818

Keywords

artificial intelligence in nephrology

data science kidney care

CRRT ICU prediction

XGBoost temporal model

TSFRESH EHR feature extraction

SHAP interpretability

hemodialysis accelerometer phenotyping

UMAP k-means clustering

CKD progression GRU model

kidney transplant decision support

computational pathology Remuzzi scoring

plasma proteomics glomerular disease classification

glomerulus-on-a-chip iPSC modeling