Abstract: SA-PO0565
Feasibility of Artificial Intelligence (AI)-Enhanced Three-Dimensional (3D) Ultrasonography for Pediatric ADPKD: A Pilot Study
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
- Gregory, Adriana, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Khalifa, Muhammed, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Bozkurt, Elif G, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Yang, Hana, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Helland, Ryan, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Irazabal, Maria V., Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Harris, Peter C., Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Kline, Timothy L., Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
- Hanna, Christian, Mayo Foundation for Medical Education and Research, Rochester, Minnesota, United States
Background
Ultrasound (US) plays a critical role in the evaluation and management of ADPKD in pediatric populations. As a non-invasive, radiation-free, and readily accessible imaging modality, it enables the detection of kidney cysts, assessment of kidney size, and monitoring of disease progression without the need for sedation. In this project, we aim to automate the quantification of total kidney volume (TKV) and total cyst volume (TCV) from 3D US images and evaluate their accuracy compared to results from MR images (the reference standard). Additionally, we analyzed the association between MR-based total cyst number (TCN) and genetic variants.
Methods
Using a Philips EPIQ scanner (C5-1 probe) and 1.5T/3T MR scanners, we acquired 3D ultrasound and MR images for automated kidney and cyst segmentation. A deep convolutional network was trained on retrospective 3D US images (n=130) from adult ADPKD subjects and evaluated on prospective data (n=40). Training used 5-fold cross-validation, stratified by patient and kidney volume (80/20 split). We compared TKV and TCV from manual and automated 3D US with automated MR measurements using linear regression and Bland–Altman analysis.
Results
We analyzed 20 patients with genetically confirmed ADPKD carrying either PKD1 or PKD2 pathogenic variants. A significant correlation was observed between the MR-based TKV and both manual and automated 3D US volumes (r=0.99, P<0.001 and r=0.85, P=0.001, respectively) and between MR-based TCV and semi-automated 3D US (r=0.92, P<0.001). Patients with a truncating PKD1 variants had more cysts (average TCN = 148) compared to patients with non-truncating PKD1, and PKD2 variants (average TCN=50) p=0.0295. Figure 1 shows MR imaging and 3D US segmentation examples for kidney and cysts.
Conclusion
This study demonstrates that the calculation of TKV and TCV are possible using 3D US images. Furthermore, advanced imaging biomarkers may offer valuable insights into cyst burden and disease prognosis in children.
Funding
- Private Foundation Support