Abstract: SA-PO0555
Artificial Intelligence in ADPKD: Transforming Diagnosis, Risk Stratification, and Therapeutic Interventions
Session Information
- Cystic Kidney Diseases: Clinical Research
November 08, 2025 | Location: Exhibit Hall, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Genetic Diseases of the Kidneys
- 1201 Genetic Diseases of the Kidneys: Monogenic Kidney Diseases
Authors
- Ebrahimi, Niloufar, Loma Linda University Medical Center, Loma Linda, California, United States
- Cheungpasitporn, Wisit, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Abdi Pour, Amir, Loma Linda University Medical Center, Loma Linda, California, United States
Background
Autosomal dominant polycystic kidney disease (ADPKD) is the most common genetic kidney disorder, leading to progressive cyst formation and kidney failure. Conventional monitoring primarily relies on imaging and eGFR trends, which may lack sensitivity in detecting early disease progression. Artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), offers transformative potential in redefining diagnosis, risk stratification, and treatment optimization in ADPKD.
Methods
A systematic review of all PubMed-indexed literature from inception to 2025 was conducted using search terms related to AI and ADPKD. Peer-reviewed publications were evaluated based on clinical relevance, AI methodologies, validation strategies, and potential clinical impact. Inclusion criteria encompassed studies with robust validation datasets, defined outcome measures, and clinical applicability. AI applications were categorized into diagnostic, prognostic, and therapeutic domains, with a focus on emerging models and their clinical implications.
Results
AI-enabled tools, particularly DL models such as U-Net and convolutional neural networks (CNNs), have demonstrated high performance in automating total kidney volume (TKV) segmentation, reducing interpretation time and inter-reader variability. ML models integrating imaging, genetic, and clinical data outperform current prognostic scores (e.g., MIC, PROPKD) in predicting eGFR decline. AI also facilitates drug repurposing, genotype-phenotype correlation, remote monitoring, and clinical trial optimization. A summary of validated AI models is provided in Table 1.
Conclusion
AI offers transformative potential in ADPKD care by improving diagnostic accuracy, predicting disease progression, and enabling precision nephrology. Future directions include multi-omics integration, real-time remote monitoring, and deployment in clinical workflows.
Table 1. Summary of AI Applications in ADPKD
| AI Modality | Application Domain | Validation Type | Performance Metric | Clinical Utility |
| U-net (nnUNet) | Diagnosis (TKV Segmentation) | DSC 0.96, TKV error 1.8% | DSC 0.96, TKV error 1.8% | Reduces manual segmentation time from 56 to 8.5 minutes |
| U-net + EfficientNet | Diagnosis (MRI TKV Automation) | DSC 0.97–0.98, Time reduction 51% | DSC 0.97–0.98, Time reduction 51% | Reduces expert time, enables scalable use |
| U-net (Axial/Coronal) | Diagnosis (Axial/Coronal TKV) | DSC 0.89 (Axial), 0.82 (Coronal) | DSC 0.89 (Axial), 0.82 (Coronal) | Comparable to experts, axial images more accurate |
| CNN | Diagnosis (Phenotype Classification) | Accuracy 98.01% | Accuracy 98.01% | Automates typical vs. atypical classification |
| XGBoost | Prognosis (Risk Prediction) | AUC 0.96 (CKD ≥3A/3B), r = 0.81 (eGFR decline) | AUC 0.96 (CKD ≥3A/3B), r = 0.81 (eGFR decline) | Stratifies patients by progression risk |
| DL Ensemble (MRI biomarkers) | Prognosis (Progression Tracking) | Fast TKV growth >5%, multiorgan imaging biomarkers | Fast TKV growth >5%, multiorgan imaging biomarkers | Tracks TKV, liver volume, and more over time |