Neurovascular Imaging with Ultra-High-Resolution Photon-Counting CT: Preliminary Findings on Image-Quality Evaluation

Study Population

This prospective, single-center study was approved by the institutional review board of the Medical University of South Carolina, and written informed consent was obtained from all participants. We enrolled consecutive patients between July and September 2023 who met the following criteria: 18 years of age or older, prior clinically-indicated standard of care (SOC) CTA of the head and neck on EID-CT within the previous 8 months, and 1 or multiple untreated intracranial saccular aneurysms. Exclusion criteria included a history of iodine contrast allergy, renal insufficiency, pregnancy, nondiagnostic SOC CTA, and scheduled neurovascular surgery (endovascular coiling or clipping). After the initial screening, the senior author reviewed each CT angiogram to confirm eligibility for the study. If it was unclear whether the patient had a saccular aneurysm or an infundibulum, then the patient was excluded from the study.

PCD-CT Image Acquisition and Reconstruction

Patients underwent head and neck CTA acquired on a clinical first-generation PCD-CT scanner (NAEOTOM Alpha; Siemens) operated in Quantum HD UHR mode, resulting in a collimation of 120 × 0.2 mm at the detector level and a reconstructed section thickness of 0.2 mm. The reconstructed matrix size was 512 × 512, and the field of view was adjusted for each patient to optimally image the vessels from the aortic arch to the vertex. We used the following acquisition parameters: tube voltage = 140 kV(peak), pitch = 0.65, rotation time = 0.5 seconds. To achieve a comparable absorbed radiation dose between scans, we used the CT dose index volume (CTDI_vol) from the EID-CT scan of each patient as a reference. The tube current–time product (milliampere-seconds) of the PCD-CT scan was adjusted to match the reference CTDI_vol.

CTA was performed after the administration of 80 mL of iodinated contrast material (iohexol, Omnipaque 350; GE Healthcare), injected through a 20-ga IV antecubital vein catheter using a power injector. The flow rate was matched to the EID-CT scan per patient, with an average of 4.37 (SD, 0.5) mL/s. Opacification of the aortic arch was monitored using a bolus-tracking technique with an attenuation threshold of 155 HU for all examinations. The start time of data acquisition was determined with a fixed delay of 5 seconds after the attenuation threshold was reached.

Axial plane, polyenergetic images (referred to as T3D by the manufacturer) were reconstructed at 0.2-mm section thickness with a 0.2-mm section increment. Quantum iterative reconstruction (referred to as QIR by the manufacturer) level 1 was used for all PCD-CT images. Four image sets were reconstructed using the following vascular kernels: body vascular [Bv]36 (least sharp, representing the standard clinical reconstruction kernel used for head and neck CTA in our institution), Bv40, Bv44, and Bv48 (Online Supplemental Data).

Quantitative Analysis

Quantitative image analysis was performed by 1 author (A.T.) using a manufacturer-specific workstation (syngo.via software version VB30; Siemens). ROIs were manually placed on 1 PCD-CT reconstruction (Bv36) at 12 standard anatomic locations, bilaterally. ROIs were copied and pasted onto the other 3 reconstructions, placing the ROIs at the exact same location with the exact same size on every PCD-CT reconstruction. The ROIs on the EID-CT images were closely matched (manually) in both size and location to the PCD-CT image sets on a per-patient basis. The data analysis used the average of the 2 bilateral ROIs at each location.

Attenuation was measured in 3 extracranial vessels: the common carotid artery, the cervical ICA (C1), and the cervical vertebral artery (V2). We selected 7 intracranial locations: the petrous segment of the ICA (C2), the carotid terminus, MCA (M1), anterior cerebral artery (A2), posterior cerebral artery (P2), intradural vertebral artery (V4), and the basilar artery. The size of the ROIs was as large as possible, ensuring that only the contrasted lumen of the artery was measured. Intracranial vessel measurements were divided into 2 groups based on the size of the ROIs: The first comprised large intracranial vessels including C2 and carotid terminus (mean ROI area = 4.28 [SD, 2.02] and 2.39 [SD, 0.74] mm², respectively), while the second consisted of small intracranial vessels such as M1, A2, P2, V4, and the basilar artery (mean ROI area ranged from 0.70 [SD, 0.20] mm² to 1.53 [SD, 0.36] mm²). Results from the analysis of the combined groups were reported.

Signal was defined as the average density (Hounsfield units); and noise, as the SD of density within the voxels of the ROIs. For the calculation of the SNR and contrast-to-noise ratio (CNR), the muscle density in the pterygoid muscle and the SD of the air adjacent to the neurocranium were measured (both ROI areas: 0.25 cm²). The SNR and CNR were calculated as follows:

Additionally, images were transferred to a dedicated workstation (ImageJ software, Version 1.53; National Institutes of Health) to analyze the sharpness of vessel borders.^17,19,20 For each image, a total of 6 line profiles were placed perpendicular to the border of the vessels to detect attenuation values at the following locations: C1 and V2 (representing the extracranial arteries), C2 and carotid terminus (representing the large intracranial arteries), M1 and A2 (representing the small intracranial arteries). For each line, the maximum per-pixel change in signal intensity (maximum ΔHU) and the full width at half maximum was calculated. Vessel sharpness was defined as the mean maximum change of CT values.

Qualitative Analysis

Qualitative image quality was assessed by 3 experienced neuroradiologists (J.A.C., M.G.M., and M.Y.). Image analysis was performed using the in-house PACS system. A training session was conducted with the readers before the image rating, and reference images were provided to ensure consistency and facilitate the application of the rating scale. The readers independently rated the images at the following intracranial locations: right and left ICA (C2 segment), PICA, vertebral artery–basilar artery junctions (VB), ophthalmic arteries (OA), carotid terminus, posterior communicating artery (PcomA), MCA trifurcation, anterior communicating artery (AcomA), basilar artery apex, and at the location of the aneurysm/s. In the first part of the evaluation, PCD-CT reconstructions and EID-CT images were displayed one by one sequentially in a randomized order, while readers were asked to rate the image quality using a 5-point Likert-type scale (Online Supplemental Data). Images were initially shown at predefined window settings (width: 1500 HU; level: 400 HU), but the readers were allowed to perform manual adjustments. Next, the 4 PCD-CT reconstructions of each patient were displayed on 1 monitor in a randomized order, and readers were asked to choose the most suitable for the evaluation of the intracranial aneurysm/s. Readers were blinded to the type of reconstructions and scanner type at each step of the image evaluation.

Statistical Analysis

Patient and examination characteristics were summarized descriptively. The normality of distributions was assessed using histograms and the Shapiro-Wilk test. The 2-sided Wilcoxon signed-rank test was used to assess differences in the distribution of qualitative and non-normally distributed quantitative image-quality scores. Scores from the qualitative image-quality analysis were pooled across readers. The interreader agreement of qualitative scores among the 3 readers was quantified with Krippendorf α coefficients (α = 0.0–0.20, poor agreement; α = 0.21–0.40, fair agreement; α = 0.41–0.60, moderate agreement; α = 0.61–0.80, substantial agreement; α = 0.81–1.00, almost perfect agreement). All P values were corrected for multiple testing using the Bonferroni correction. Two-tailed P values < .05 were considered statistically significant. If not stated otherwise, all data are presented as mean (SD). All statistical analyses were performed by using SPSS, Version 28.0.1 (IBM).