Coronary calcium scoring using virtual non-contrast reconstructions on a dual-layer spectral CT system: Feasibility in the clinical practice

Virtual non-contrast series of photon-counting detector computed tomography angiography for aortic valve calcium scoring

The aim of our study was to evaluate two different virtual non-contrast (VNC) algorithms applied to photon counting detector (PCD)-CT data in terms of noise, effectiveness of contrast media subtraction and aortic valve calcium (AVC) scoring compared to reference true non-contrast (TNC)-based results. Consecutive patients underwent TAVR planning examination comprising a TNC scan, followed by a CTA of the heart. VNC series were reconstructed using a conventional (VNCconv) and a calcium-preserving (VNCpc) algorithm. Noise was analyzed by means of the standard deviation of CT-values within the left ventricle. To assess the effectiveness of contrast media removal, heart volumes were segmented and the proportion of their histograms > 130HU was taken. AVC was measured by Agatston and volume score. 41 patients were included. Comparable noise levels to TNC were achieved with all VNC reconstructions. Contrast media was effectively virtually removed (proportions > 130HU from 81% to < 1%). Median calcium scores derived from VNCconv underestimated TNC-based scores (up to 74%). Results with smallest absolute difference to TNC were obtained with VNCpc reconstructions (0.4 mm, Br36, QIR 4), but with persistent significant underestimation (median 29%). Both VNC algorithms showed near-perfect (r²>0.9) correlation with TNC. Thin-slice VNC reconstructions provide equivalent noise levels to standard thick-slice TNC series and effective virtual removal of iodinated contrast. AVC scoring was feasible on both VNC series, showing near-perfect correlation, but with significant underestimation. VNCpc with 0.4 mm slices and Br36 kernel at QIR 4 gave the most comparable results and, with further advances, could be a promising replacement for additional TNC.

Heart rate sensitivity of virtual non-contrast calcium scores derived from photon counting detector CT data: a phantom study

PURPOSE

To assess the reliability of virtual non-contrast (VNC) derived coronary artery calcium quantities in relation to heart rate and the VNC algorithm used compared to reference true non-contrast (TNC), considering several clinically established acquisition modes.

MATERIAL AND METHODS

An ad hoc built coronary phantom containing four calcified lesions and an iodinated lumen was scanned using three cardiac acquisition modes three times within an anthropomorphic cardiac motion phantom simulating different heart rates (0, 60, 80, 100 bpm) and reconstructed with a conventional (VNCconv) and a calcium-sensitive (VNCpc) VNC algorithm. TNC reference was scanned at 0 bpm with non-iodinated lumen. Calcium scores were assessed in terms of number of lesions detected, Agatston and volume scores and global noise was measured. Paired t-test and Wilcoxon test were performed to test measurements for significant difference.

RESULTS

For both VNC algorithms used, calcium levels or noise were not significantly affected by heart rate. Measurements on VNCpc reconstructions best reproduced TNC results, but with increased variability (Agatston scores at 0 bpm for TNC, VNCconv, and VNCpc were 47.1 ± 1.1, 6.7 ± 2.8 (p < 0.001), and 45.3 ± 7.6 (p > 0.05), respectively). VNC reconstructions showed lower noise levels compared to TNC, especially for VNCpc (noiseheart on TNC, VNCconv and VNCpc at 0 bpm was 5.0 ± 0.4, 4.5 ± 0.2, 4.2 ± 0.2).

CONCLUSION

No significant heart rate dependence of VNC-based calcium scores was observed in an intra-reconstruction comparison. VNCpc reproduces TNC scores better than VNCconv without significant differences and decreased noise, however, with an increasing average deviation with rising heart rates. VNC-based CACS should be used with caution as the measures show higher variability compared to reference TNC and therefore hold the potential of incorrect risk categorization.

Coronary Artery Calcium Scoring Using Virtual Versus True Noncontrast Images From Photon-Counting Coronary CT Angiography

Background Coronary artery calcium scoring (CACS) for coronary artery disease requires true noncontrast (TNC) CT alongside contrast-enhanced coronary CT angiography (CCTA). Photon-counting CT provides an algorithm (PureCalcium) for reconstructing virtual noncontrast images from CCTA specifically for CACS. Purpose To assess CACS differences based on PureCalcium images derived from contrast-enhanced photon-counting CCTA compared with TNC images and evaluate the impact of these differences on the clinically relevant classification of patients into plaque burden groups. Materials and Methods Photon-counting CCTA images acquired between August 2022 and May 2023 were retrospectively identified. Agatston scores were derived from both TNC and PureCalcium images and tested for differences with use of the Wilcoxon signed-rank test. The agreement was assessed with use of equivalence tests, Bland-Altman analysis, and intraclass correlation coefficient. Plaque burden groups were established based on Agatston scores, and agreement was evaluated using weighted Cohen kappa. The dose-length product was analyzed. Results Among 170 patients (mean age, 63 years ± 13 [SD]; 92 male), 111 had Agatston scores higher than 0. Median Agatston scores did not differ between TNC and PureCalcium images (4.8 [IQR, 0-84.4; range, 0.0-2151.8] vs 2.7 [IQR, 0-90.7; range, 0.0-2377.1]; P = .99), with strong correlation (intraclass correlation coefficient, 0.98 [95% CI: 0.97, 0.99]). The equivalence test was inconclusive, with a 95% CI of 0.90, 1.19. Bland-Altman analysis showed wide repeatability limits, indicating low agreement between the two scores. With use of the PureCalcium algorithm, 125 of 170 patients (74%) were correctly classified into plaque burden groups (excellent agreement, κ = 0.88). Patients without plaque burden were misclassified at higher than normal rates (P < .001). TNC image acquisition contributed a mean of 19.7% ± 8.8 of the radiation dose of the entire examination. Conclusion PureCalcium images show potential to replace TNC images for measuring Agatston scores, thereby reducing radiation dose in CCTA. There was strong correlation in calcium scores between TNC and PureCalcium, but limited agreement. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Sakuma in this issue.

Diagnostic accuracy of virtual non-contrast CT for aortic valve stenosis severity evaluation

BACKGROUND

Computed tomography aortic valve calcium (AVC) score has accepted value for diagnosing and predicting outcomes in aortic stenosis (AS). Multi-energy CT (MECT) allows virtual non-contrast (VNC) reconstructions from contrast scans. We aim to compare the VNC-AVC score to the true non-contrast (TNC)-AVC score for assessing AS severity.

METHODS

We prospectively included patients undergoing a MECT for transcatheter aortic valve replacement (TAVR) planning. TNC-AVC was acquired before contrast, and VNC-AVC was derived from a retrospectively gated contrast-enhanced scan. The Agatston scoring method was used for quantification, and linear regression analysis to derive adjusted-VNC values.

RESULTS

Among 109 patients (55% female) included, 43% had concordant severe and 14% concordant moderate AS. TNC scan median dose-length product was 116 ​mGy∗cm. The median TNC-AVC was 2,107 AU (1,093-3,372), while VNC-AVC was 1,835 AU (1293-2,972) after applying the coefficient (1.46) and constant (743) terms. A strong correlation was demonstrated between methods (r ​= ​0.93; p ​< ​0.001). Using accepted thresholds (>1,300 AU for women and >2,000 AU for men), 65% (n ​= ​71) of patients had severe AS by TNC-AVC and 67% (n ​= ​73) by adjusted-VNC-AVC. After estimating thresholds for adjusted-VNC (>1,564 AU for women and >2,375 AU for men), 56% (n ​= ​61) had severe AS, demonstrating substantial agreement with TNC-AVC (κ ​= ​0.77).

CONCLUSIONS

MECT-derived VNC-AVC showed a strong correlation with TNC-AVC. After adjustment, VNC-AVC demonstrated substantial agreement with TNC-AVC, potentially eliminating the requirement for an additional scan and enabling reductions in both radiation exposure and acquisition time.

Image Characteristics of Virtual Non-Contrast Series Derived from Photon-Counting Detector Coronary CT Angiography-Prerequisites for and Feasibility of Calcium Quantification

In photon-counting detector CT (PCD-CT), coronary artery calcium scoring (CACS) can be performed using virtual non-contrast (VNC) series derived from coronary CT angiography (CCTA) datasets. Our study analyzed image characteristics of VNC series in terms of the efficacy of virtual iodine "removal" and image noise to determine whether the prerequisites for calcium quantification were satisfied. We analyzed 38 patients who had undergone non-enhanced CT followed by CCTA on a PCD-CT. VNC reconstructions were performed at different settings and algorithms (conventional VNCConv; PureCalcium VNCPC). Virtual iodine "removal" was investigated by comparing histograms of heart volumes. Noise was assessed within the left ventricular cavity. Calcium was quantified on the true non-contrast (TNC) and all VNC series. The histograms were comparable for TNC and all VNC. Image noise between TNC and all VNC differed slightly but significantly. VNCConv CACS showed a significant underestimation regardless of the reconstruction setting, while VNCPC CACS were comparable to TNC. Correlations between TNC and VNC were excellent, with a higher predictive accuracy for VNCPC. In conclusion, the iodine contrast can be effectively subtracted from CCTA datasets. The remaining VNC series satisfy the requirements for CACS, yielding results with excellent correlation compared to TNC-based CACS and high predicting accuracy.

Calcium scoring on coronary computed angiography tomography with photon-counting detector technology: Predictors of performance

INTRODUCTION

Obtaining accurate coronary artery calcium (CAC) score measurements from CCTA datasets with virtual non-iodine (VNI) algorithms would reduce acquisition time and radiation dose. We aimed to assess the agreement of VNI-derived and conventional true non-contrast (TNC)-based CAC scores and to identify the predictors of accuracy.

METHODS

CCTA datasets were acquired with either 120 or 140 ​kVp. CAC scores and volumes were calculated from TNC and VNI images in 197 consecutive patients undergoing CCTA. CAC density score, mean volume/lesion, aortic Hounsfield units and standard deviations were then measured. Finally, percentage deviation (VNI - TNC/TNC∗100) of CTA-derived CAC scores from non-enhanced scans was calculated for each patient. Predictors (including anthropometric and acquisition parameters, as well as CAC characteristics) of the degree of discrepancy were evaluated using linear regression analysis.

RESULTS

While the agreement between TNC and VNI was substantial (mean bias, 6.6; limits of agreement, 178.5/145.3), a non-negligible proportion of patients (36/197, 18.3%) were falsely reclassified as CAC score ​= ​0 on VNI. The use of higher tube voltage significantly decreased the percentage deviation relative to TNC-based values (β ​= ​-0.21 [95%CI: 0.38 to -0.03], p ​= ​0.020) and a higher CAC density score also proved to be an independent predictor of a smaller difference (β ​= ​-0.22 [95%CI: 0.37 to -0.07], p ​= ​0.006).

CONCLUSION

The performance of VNI-based calcium scoring may be improved by increased tube voltage protocols, while the accuracy may be compromised for calcified lesions of lower density. The implementation of VNI in clinical routine, however, needs to be preceded by a solution for detecting smaller lesions as well.

Artificial intelligence in coronary artery calcium measurement: Barriers and solutions for implementation into daily practice

Coronary artery calcification (CAC) measurement is a valuable predictor of cardiovascular risk. However, its measurement can be time-consuming and complex, thus driving the desire for artificial intelligence (AI)-based approaches. The aim of this review is to explore the current status of CAC volume measurement using AI-based systems for the automated prediction of cardiovascular events. We also make proposals for the implementation of these systems into clinical practice. Research to date on applying AI to CAC scoring has shown the potential for automation and risk stratification, and, overall, efficacy and a high level of agreement with categorisation by trained clinicians have been demonstrated. However, research in this field has not been uniform or directed. One contributing factor may be a lack of integration and communication between computer scientists and cardiologists. Clinicians, institutions, and organisations should work together towards applying this technology to improve processes, preserve healthcare resources, and improve patient outcomes.