Coronary plaque quantification on energy-integrating and photon-counting detector CT: reproducibility and power modeling for mixed-platform trials.
Authors
Affiliations (8)
Affiliations (8)
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC, 29425, USA.
- Heart and Vascular Centre, Semmelweis University, 68 Varosmajor Street, Budapest, 1122, Hungary.
- Siemens Healthineers, 755 College Road E, Princeton, NJ, 08540, USA.
- Department of Radiology, Medical Imaging Centre, Semmelweis University, 78a Ulloi Avenue, Budapest, 1082, Hungary.
- Department of Radiology and Radiological Science, Medical University of South Carolina, 25 Courtenay Drive, Charleston, SC, 29425, USA. [email protected].
- Department of Radiology, Medical Imaging Centre, Semmelweis University, 78a Ulloi Avenue, Budapest, 1082, Hungary. [email protected].
- Department of Diagnostic and Interventional Radiology, University Medical Center of the Johannes Gutenberg-University, 1 Langenbeckstraße, 55131, Mainz, Germany.
- German Centre for Cardiovascular Research, Partner Site Rhine-Main, 1 Langenbeckstraße, 55131, Mainz, Germany.
Abstract
To evaluate the reproducibility of quantitative plaque characterization between energy-integrating detector (EID)-CT and photon-counting detector (PCD)-CT, identify reconstruction settings yielding the lowest variability, and model sample-size requirements for trial planning. Patients who underwent coronary CT angiography on dual-source EID-CT and PCD-CT within 30 days were screened retrospectively. EID-CT data were reconstructed using quantitative (Qr40) and vascular (Bv40) kernels, while PCD-CT data were reconstructed with Bv36/40/44 and Qr36/40/44 kernels and quantum iterative reconstruction strengths 2-4. For each plaque component (total, low-attenuation, fibrotic, and calcified), volumes were computed using fixed and adaptive Hounsfield-unit thresholds using an automated deep-learning-based plaque-quantification platform and spatially co-registered. Inter-scanner standard deviation (SD) was calculated, and optimal reconstruction pairs were used for power modeling. Thirty-eight patients (age 68.0 [64.0-72.5] years, 30 men) and 77 vessels were included. Inter-scanner correlations were very strong for total (r = 0.85-0.95), fibrotic (r = 0.74-0.94), and calcified plaque (r = 0.92-0.98), and strong for low-attenuation plaque volumes (r = 0.71-0.85). Mean bias ranged from 8.2 to 164.4 mm³ for total plaque volume. Applying the optimal reconstruction pair (Qr40<sub>EID-CT</sub> vs. Bv36<sub>PCD-CT</sub>), inter-scanner SDs were 0.14 (vessel-based) and 0.12 (patient-based). At 80% power and α = 0.05, the estimated sample sizes to detect 5% and 10% changes in total plaque volume were 116 and 29 vessels or 92 and 23 patients per group. Quantitative coronary plaque volumes demonstrated high inter-scanner consistency between EID-CT and PCD-CT under optimized conditions. The dual-level power-modeling framework translates inter-scanner variability into actionable sample-size estimates for study design. Question Quantitative coronary plaque volumes are increasingly used as surrogate biomarkers, yet variability between energy-integrating and photon-counting CT remains a major concern. Findings Harmonized reconstruction settings achieved high inter-scanner consistency with low variability when optimized reconstruction settings and automated plaque quantification are applied. Clinical relevance Standardized coronary plaque quantification across CT technologies enables reliable use of quantitative plaque endpoints, supporting mixed-platform longitudinal trials.