Photon-counting CT-enabled urolithiasis phenotyping: virtual non-iodine reconstructions, automated measurement, and spectral radiomics.
Authors
Affiliations (2)
Affiliations (2)
- Urology Department, Institute of Medical Sciences, Faculty of Medicine, Collegium Medicum, Cardinal Stefan Wyszyński University in Warsaw, Warsaw, Poland. [email protected].
- Department of Urology and Urological Oncology, Multidisciplinary Hospital in Warsaw-Miedzylesie, Warsaw, Poland. [email protected].
Abstract
Urolithiasis management increasingly depends on accurate, noninvasive stone phenotyping to guide acute intervention, secondary prevention, and selective chemolitholysis. Photon-counting computed tomography (PCCT) introduces detector-level energy discrimination and higher spatial resolution, enabling calcium-preserving reconstruction strategies and quantitative spectral analytics that may shift stone characterization from a laboratory endpoint toward an imaging-derived biomarker. Recent peer-reviewed PCCT studies have concentrated on three translational domains. First, calcium-preserving virtual non-iodine (VNI) and virtual non-contrast (VNC) reconstructions have been evaluated for upper-tract stone detection in contrast-enhanced settings, supporting the concept that a single contrast-enhanced acquisition could potentially replace multiphase protocols in selected scenarios. Second, comparative ex vivo and clinical imaging studies suggest that PCCT improves depiction of small calculi and enables automated, high-resolution stone burden quantification. Third, spectral radiomics and machine-learning models have been applied to monoenergetic PCCT reconstructions for multi-class stone composition discrimination, achieving high discriminatory performance in controlled ex vivo datasets, and complementary phantom work has demonstrated automated uric acid versus non-uric acid classification. The emerging literature suggests that PCCT may support calcium-preserving assessment of stones in contrast-enhanced imaging, automated and reproducible stone burden quantification, and composition phenotyping via spectral analytics. However, most studies remain phantom/ex vivo and highly platform-specific. Translation will depend on prospectively defined acquisition and reconstruction parameters, externally validated models, and rigorous reporting aligned with contemporary machine-learning standards.