Artificial Intelligence (AI)-Driven Estimation of Total Kidney Volume (TKV) in ADPKD

Artificial Intelligence (AI)-Driven Estimation of Total Kidney Volume (TKV) in ADPKD

This poster, presented at the American Society of Neurology (ASN) 2025 Kidney Week meeting reviews the development and validation of artificial intelligence (AI)-driven models for automated estimation of Total Kidney Volume (TKV) in patients with Autosomal Dominant Polycystic Kidney Disease (ADPKD) using MRI scans. By leveraging a retrospective cohort from the CRISP study and public datasets, the team compared 2D, 3D, and ensemble deep learning models, demonstrating that expanding the training dataset and ensemble modeling significantly improved segmentation accuracy and boundary precision. The results show that AI-based tools can reliably and reproducibly estimate TKV, supporting their feasibility for use as prognostic biomarkers in clinical trials and routine research settings. 

Why download this poster?

• Regulatory Relevance: Demonstrates alignment with FDA guidance on imaging-based TKV as a qualified prognostic biomarker in ADPKD trials.
• AI Innovation: Showcases advanced AI models (2D, 3D, ensemble) for automated, reproducible kidney segmentation and volume estimation.
• Clinical Impact: Provides evidence that AI-driven methods improve accuracy and generalizability, supporting robust endpoint measurement in clinical trials.
• Operational Efficiency: Highlights how automation can reduce manual workload, increase repeatability, and streamline imaging workflows.
• Data-Driven Insights: Includes detailed performance metrics (DSC, IoU, Hausdorff distance, volume error) and visual examples to inform decision-making.
• Translational Value: Offers practical insights for deploying AI tools in both research and clinical practice, accelerating drug development for kidney diseases.