Patients

In this retrospective, single-center study, consecutive patients with severe aortic stenosis undergoing a PCD-CT, which comprised a CCTA and a cardiac LE scan between May 2022 and November 2023 were screened. Scans were acquired as part of the diagnostic work-up for TAVR planning. Exclusion criteria were missing raw data, prior valve replacement, coronary stenting, aortocoronary bypass grafting, or pacemaker leads as foreign material may impede measurements, severe streak artifacts from gross calcifications, and missing blood hematocrit. Studies with an ultra-high-resolution CCTA were also excluded. The first 75 patients were used as a derivation cohort. The following 50 patients served as a validation cohort (Fig. 1).

Fig. 1

Flowchart detailing patient inclusion. CCTA, coronary CT angiography; PCD-CT, photon-counting detector CT; TAVR, transcatheter aortic valve replacement

All patients provided written informed consent to allow the inclusion of their anonymized data in retrospective analyses. Institutional review board and local ethics committee approval were obtained.

CT data acquisition and image reconstruction

Scans were performed on a first-generation dual-source PCD-CT system (NAEOTOM Alpha; Siemens Healthineers AG) equipped with two cadmium telluride detectors. The scan protocol started with an electrocardiography (ECG)-gated non-contrast cardiac scan, followed by an ECG-gated CCTA, a whole-body aortography, and ending with an ECG-gated cardiac LE scan. CCTA was initiated after the injection of a weight-based volume of iodinated contrast medium (60–80 mL, iopromide, Ultravist 370 mgI/mL; Bayer Healthcare) accompanied by a saline chaser (20 mL, NaCl 0.9%) into an antecubital vein applying a weight-based flow rate (3.3–4.4 mL/s). Cardiac LE scans were acquired five minutes after contrast media injection.

The ECG-gated sequential multi-energy QuantumPlus mode with a detector collimation of 144 × 0.4 mm was applied for CCTA and LE scans. CCTA and LE scans were acquired with a tube potential of 140 kV and 120 kV, respectively, both using automated tube current modulation (CAREDose4D, Siemens) with an image quality level of 64 and 80, respectively. The ECG-pulsing window was fixed from 30% to 60% of the R-R interval for the CCTA and to an absolute interval of 280 milliseconds from the R wave for the LE scan. The gantry rotation time was 0.25 s. No beta-blockers were administered.

Derivation cohort—image reconstruction and linear regression analysis

CCTA and cardiac LE scans were reconstructed as calcium-preserving VNI images at 60, 70, and 80 keV [14, 24] as well as VNC images [22, 25] applying quantum iterative reconstruction (QIR) strengths 2, 3, and 4 [26], resulting in a total of 24 reconstructions for every patient (Fig. 2). VNI and VNC images were reconstructed with a slice thickness of 3 mm and increment of 1.5 mm using the quantitative Qr36 kernel. All images were reconstructed with a field of view of 200 × 200 mm2 and using a matrix size of 512 × 512 pixels.

Fig. 2

Representative CT images obtained in an 85-year-old man with severe aortic stenosis examined for transcatheter aortic valve replacement planning. Axial CT images acquired with dual-source photon-counting detector CT show the virtual non-iodine (VNI) images at 60, 70, and 80 keV and the virtual non-contrast (VNC) images reconstructed from coronary CT angiography and cardiac late enhancement scan with quantum iterative reconstruction (QIR) strengths 2, 3, and 4

One reader (N.E.), who was blinded to clinical information including the blood hematocrit, measured the CT attenuation of the blood pool by placing circular regions of interest (ROIs) in the left ventricle, and in the ascending and descending aorta at the level of the right pulmonary artery. The ROIs were first positioned in CCTA with the largest possible diameter while carefully avoiding adjacent structures and were then copied to VNI and VNC images from CCTA and LE scans. All image sets were linked to show identical anatomical structures. Mean blood pool attenuation (BPmean) was calculated by averaging the attenuation values of all three regions. BPmean of every reconstruction was compared with the blood hematocrit with linear regression analyses and Pearson correlation.

Validation cohort—image reconstruction, synthetic hematocrit calculation, and extracellular volume computation

Cardiac LE scans were reconstructed as VNC images applying QIR 3 according to the results of the derivation cohort and supported by previous studies [22, 23] to calculate the synthetic hematocrit. Moreover, specific images were reconstructed from CCTA and cardiac LE scans to allow for the computation of the myocardial ECV: virtual monoenergetic images (VMI) at 65 keV from CCTA and LE scans, and iodine images from LE scans. The ECV-specific images were reconstructed with a slice thickness of 1.5 mm and increment of 1 mm, using the quantitative Qr40 kernel and applying QIR 3 as detailed in previous studies [20, 21]. All images were reconstructed with a field of view of 200 × 200 mm2 and using a matrix size of 512 × 512 pixels. Supplemental Fig. 1 delineates the various reconstructed images and their corresponding tasks.

The same reader (N.E.) measured BPmean in a single image set. The synthetic hematocrit was calculated using BPmean obtained as described above, employing the corresponding linear regression formula determined in the derivation cohort. A second reader (V.M.) computed the myocardial ECV with a semi-automatic software (Cardiac Functional Analysis Frontier Version 2.1.0, Syngo.via, Siemens Healthineers AG) applying the iodine image-derived iodine concentration in the myocardium and the blood pool of the LE scan [3]:

$$}}}}}\; }}}}}}=\left(1-}}}}}}\right)* \frac}}}}}}}_}}}}}}}]}}}}}}}}_}}}}}\; }}}}}}}]}$$

For myocardial ECV calculation, either the blood hematocrit (determined via blood drawing) or a synthetic hematocrit (derived from ROI measurements directly on the images as described above) served as input values. Mean midmyocardial ECV, which included the myocardium from the inner 25% to the outer 25%, was noted.

Statistical analysis

Analyses were performed using R statistical software (R, version 4.3.2; R Foundation, https://www.R-project.org/). Variables were tested for normal distribution with the Shapiro-Wilk test. Continuous variables were summarized as means and standard deviations or medians and interquartile ranges when normally or nonnormally distributed. Categorical variables were presented as counts and percentages. In the derivation cohort, BPmean was correlated to the blood hematocrit with linear regression analyses and Pearson correlations [27]. In the validation cohort, myocardial ECV using the blood hematocrit and the synthetic hematocrit were compared with Bland-Altman analyses and Wilcoxon signed-rank tests. Patient characteristics and blood hematocrit between the derivation and validation cohorts were compared using t-tests and the Wilcoxon signed-rank tests, as appropriate. A two-tailed p-value less than 0.05 was considered to indicate statistical significance.