Photon-counting detector of the ophthalmic artery: impact of virtual monoenergetic levels, reconstruction kernel, and quantum iterative reconstruction (QIR) on image quality

Introduction

The ophthalmic artery (OA), the first major branch of the internal carotid artery, supplies blood to the orbit and its critical structures, including the retina and optic nerve (1). Accurate visualization of the OA is essential for diagnosing various ocular conditions, including ischemic optic neuropathies, ocular ischemic syndrome, orbital tumors, and evaluating collateral flow in cerebrovascular disease. Further, alterations in OA flow dynamics are increasingly recognized as potential biomarkers in systemic diseases such as diabetes mellitus, obesity and hypertension (2-4).

However, visualizing the OA presents significant challenges. The complex anatomical characteristics of the OA, such as its small diameter (approximately 1.5 mm) (5-7), numerous branches, tortuous course within the orbit, and proximity to the bony structures at the skull base, requires high resolution, thin sections, and relatively rapid imaging to ensure high intravascular contrast. While digital subtraction angiography is considered the gold standard for detailed vascular morphology due to its superior spatial and temporal resolution (8-10), its invasiveness limits its routine application.

The introduction of photon-counting detector computed tomography (PCD-CT) into clinical practice meets the requirements for OA imaging. Farnsworth et al. demonstrated that the arterial anatomy of the orbit is much better depicted with PCD-CT compared with traditional energy-integrating detector computed tomography (EID-CT) (11), as image quality is improved through superior spatial resolution, the elimination of electronic noise, the maintenance of high temporal resolution, and the application of spectral capabilities (12-14). In addition, image contrast in PCD-CT is improved by more optimal photon weighting and the multi-energy capability.

Virtual monoenergetic imaging (VMI) at low keV levels can increase luminal iodine contrast, which is beneficial for the visualization of arterial vessels (15-19), although lower keV levels also increase image noise (20,21). In addition to the keV level, PCD-CT offers a range of adjustable parameters, including the reconstruction kernel, quantum iterative reconstruction (QIR) algorithm, slice thickness, and field of view. These parameters can significantly influence image quality, affecting spatial resolution, noise levels, and artifact suppression (22).

To date, no study has systematically assessed the effects of varying these parameters (keV level, kernel and QIR) on the clarity and diagnostic utility of OA visualization in PCD-CT. Therefore, this study aimed to use objective and subjective image quality assessments to determine the appropriate combination of the reconstruction kernel, keV level, and QIR to optimize the image quality of OA visualization. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2026-1-0247/rc).

Methods

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Academic Ethics Committee of Shaoxing People’s Hospital. This was a retrospective study without any intervention on participants. It did not affect the patients’ clinical diagnosis, treatment plan or outcome, so the requirement for patient approval or written informed consent was waived.

Participant population

All consecutive patients who underwent computed tomography angiography (CTA) of the carotid artery on a PCD-CT scanner between May and June 2025 were included in this retrospective study. The exclusion criteria were as follows: (I) CTA protocols not performed in ultra-high resolution (UHR) scanning mode; (II) a history of cranial surgery (tumor, trauma, or vascular malformation); (III) the presence of metal artifacts and moving artifacts affecting image quality; and/or (IV) an OA diameter <1.0 mm. The study inclusion process is displayed in the flowchart in Figure 1.

Figure 1 Flowchart showing the patient selection process, including inclusion and exclusion criteria, in this study. OA, ophthalmic artery; PCD-CT, photon-counting detector computed tomography; UHR, ultra-high resolution.

CTA acquisition and reconstruction

Patients were examined on a first-generation dual-source PCD-CT system (NAEOTOM Alpha, Siemens Healthineers, Erlangen, Germany) using the UHR scanning mode with the following parameters: tube voltage, 120 kVp; collimation, 120 mm × 0.2 mm; pitch, 0.85; and rotation time, 0.25 s. Automated tube current modulation (CARE Dose4D, Siemens Healthineers, Erlangen, Germany) was used with an image quality level of 145. This level, representing a target image noise [standard deviation (SD)] in a reference phantom, was selected based on our institutional standard protocol for head and neck CTA. The protocols reflect a balance between achieving sufficient image quality for the diagnostic assessment of small vessels such as the OA and adhering to radiation dose optimization principles, informed by vendor recommendations, initial phantom/patient testing, and clinical experience with similar EID-CT.

Patients were positioned supine with arms resting at their sides, and the head was stabilized using a cushion to minimize motion artifacts. The scan range extended from the aortic arch to the vertex of the skull. Contrast media (Xenetix 350 mgI/mL; Guerbet, Villepinte, France) followed by a saline chaser were administered via a dual-syringe power injector [computed tomography (CT) Motion, Ulrich Medical, Ulm, Germany] according to a weight-adapted protocol: For patients weighing ≤50 kg, a flow rate of 4–5 mL/s was recommended to balance vascular enhancement and minimize the risk of extravasation; for patients weighing >50 kg, a flow rate of 5–6 mL/s was recommended; rates >6 mL/s were avoided to prevent venous contamination in the orbital region. Regarding contrast volume, the standard dosage was 1.2 mL/kg body weight, with an upper limit of 100 mL to minimize renal toxicity. Additionally, a fixed 30-mL saline chaser, administered at the same flow rate as the contrast, was used to optimize bolus geometry. Image acquisition was initiated 5 seconds after a threshold of 150 Hounsfield unit (HU) was reached in the ascending aorta.

Axial VMI series were reconstructed at different virtual energy levels from 40 to 120 keV in 20 keV increments using the dedicated postprocessing software (SyngoVia, VA50 Siemens Healthineers, Erlangen, Germany). All reconstructions used body vascular convolution (Bv) kernels, with the following parameters: slice thickness, 0.4 mm; slice increment, 0.4 mm; kernels: Bv40, 48, 56, 64, and 72; matrix size, 768×768; and QIR strength, 3 and 4.

Objective analysis

Quantitative image analysis was independently performed by two board-certified neuroradiologists, blinded to reconstruction parameters, (with 10 and 15 years of experience, respectively), using 3D Slicer (version 5.6.2) to place standardized regions of interest (ROIs) across three axial slices: three circular 1.0-mm² ROIs in the intracranial, intracanicular, and intraorbital segments of the OA (23).

The intracranial, intracanalicular, and intraorbital segments of the OA extend from its origin at the supraclinoid internal carotid artery, through the optic canal adjacent to the optic nerve, and into the orbit where it terminates in its terminal branches. Mean arterial signal intensity attenuation (MA, HU) was calculated as the average of all ROI measurements; five 3-mm² ROIs in homogeneous intraorbital fat (yielding SD_fat = noise SD); three 2-mm² ROIs at medial rectus muscle belly (yielding HU_muscle = muscle attenuation). This methodology, while subject to inherent operator variability in small vessels, is consistent with established practices for quantitative CTA analysis and focuses on relative differences between reconstruction settings (24,25).

Key metrics were calculated as follows:

• MA: mean attenuation value of the OA ROIs (HU).
• Signal-to-noise ratio [SNR]: $\text{SNR}=\frac{\text{MA}}{{\text{SD}}_{\text{fat}}}$
• Contrast-to-noise ratio: $\text{CNR}=\frac{\left|\text{MA}-{\text{HU}}_{\text{muscle}}\right|}{{\text{SD}}_{\text{fat}}}$

Subjective analysis

Two board-certified radiologists (with 10 and 15 years of experience, respectively) independently evaluated the top five protocols in randomized order using a standardized protocol. The evaluators were blinded to reconstruction parameters and assessed images on a dedicated workstation with adjustable window settings (initial level/width: 700/80 HU). CT images were presented randomly, and the evaluators rated the overall image quality, sharpness of the vessel and diagnostic assess-ability. Three subjective parameters were rated using 5-point Likert scales: overall image quality (1 = non-diagnostic, 5 = excellent), vessel sharpness (1 = blurred margins, 5 = razor-sharp delineation), and diagnostic assess-ability (1 = uncertain, 5 = definitive diagnosis). Inter-reader agreement was calculated using Cohen’s kappa (κ), with consensus review for discrepant cases (κ>0.8 indicating near-perfect agreement). The composite quality score (sum of three metrics; maximum 15 points) enabled quantitative comparison across reconstruction techniques.

Statistical analysis

Descriptive statistics were computed for all study variables. For quantitative data, the normality of distribution was first assessed using the Shapiro-Wilk test. Normally distributed continuous variables are presented as the mean ± SD. Non-normally distributed continuous variables are summarized as the median with the interquartile range (IQR). For qualitative variables, data are presented as frequencies and percentages (n, %). Inter-rater reliability for objective features was assessed using the intraclass correlation coefficient (ICC) based on a two-way mixed-effects model for absolute agreement {ICC [3, 1]}. Interobserver agreement was interpreted as follows: ICC values greater than 0.90 indicated excellent agreement, values between 0.75 and 0.90 indicated good agreement, values between 0.75 and 0.50 indicated moderate agreement, and values less than 0.50 indicated fair to poor agreement. Inter-rater reliability for subjective features was assessed using Cohen’s kappa coefficients, and interpreted as follows: values between 0.81 and 1.00 indicated excellent agreement, values between 0.61 and 0.80 indicated good agreement, values between 0.41 and 0.60 indicated moderate agreement, values between 0.21 and 0.40 indicated fair agreement, and values less than or equal to 0.20 indicated poor agreement. The statistical analysis employed linear mixed-effects models (LMMs) to evaluate the impact of the scanning parameters (tube voltage, reconstruction kernel, and QIR level) on the image quality metrics (the MA, SNR, and CNR). Patient-specific variability was accounted for by incorporating random intercepts for each patient. Three separate LMMs were constructed using the lme4 package in R (version 4.3.0), with each model predicting one image quality metric as the outcome variable while including all main effects and interaction terms of the scanning parameters. Parameter optimization was performed by generating predicted values across all possible combinations of virtual energy levels (40–120 keV at 20 keV increments), reconstruction kernels (Bv40, 48, 56, 64, and 72), and QIR levels 3 and 4. To balance the three image quality metrics, a composite score was calculated by summing the z-score standardized values (mean =0, SD =1) of the predicted metrics. The top five parameter combinations were identified based on this composite score. The Friedman test was employed as a non-parametric repeated measures alternative to analysis of variance to assess global differences among the five parameter combinations. Where Friedman tests indicated significant differences (P<0.05), Dunn’s test with Bonferroni correction was applied for pairwise comparisons. All statistical analyses were performed using R version 4.2.2 (R Foundation for Statistical Computing), with a two-tailed significance threshold of α=0.05.

Results

Participants

Between May and June 2025, 67 consecutive patients (134 OAs) underwent carotid CTA using PCD-CT technology. After exclusions due to: (I) non-UHR acquisition (n=34); (II) prior cranial surgery (n=9); (III) image-compromising artifacts (n=7); and (IV) submillimeter OA diameter (n=1), 17 patients (33 OAs) qualified for analytical assessment (mean age 71.4±10.1 years; IQR, 66–80 years).

The demographic and clinical characteristics are summarized as follows: the cohort included 10 males (58.8%) and 7 females (41.2%). The median body mass index (BMI) was 23.9 kg/m² (IQR, 22.9–24.7 kg/m²). The computed tomography dose index (CTDI_vol) and dose length product (DLP) were 10.80±0.63 mGy and 333.82±40.76 mGy·cm, respectively. Cardiovascular risk factor distribution revealed hypertension in 3 patients (17.6%), high cholesterol in 4 (23.5%), and diabetes mellitus in 5 (29.4%). Current smoking was reported by 4 patients (23.5%). For further details, see Table 1.

Inter-rater reliability

Objective and subjective assessments demonstrated robust agreement between the independent readers. For the objective metrics, inter-rater reliability was excellent for the MA [ICC =0.968; 95% confidence interval (CI): 0.965–0.971]. Both the CNR (ICC =0.821; 95% CI: 0.804–0.837) and SNR (ICC =0.789; 95% CI: 0.767–0.809) showed good agreement. Subjective image quality assessments revealed excellent consistency across all domains: overall image quality (κ=0.847; P<0.001), vessel sharpness (κ =0.867; P<0.001), and diagnostic assess-ability (κ=0.868; P<0.001).

Objective analysis

The density and boxplot analysis (Figure S1) revealed distinct distribution patterns across the image quality metrics: the MA showed a right-skewed distribution (median, 456.2 HU; range, 398.5–523.8 HU); the SNR exhibited the widest variation (median, 12.1; range, 4.8–22.3); and the CNR demonstrated the most symmetric distribution (median, 9.4; range, 7.6–11.2). Table 2 sets out the objective analysis scores for the MA, SNR, and CNR of the OA.

The highest CNR was obtained at 40 keV using Bv40 and QIR4. Lower keV levels (40–60 keV) increased the MA by 136–195% but decreased the SNR by 54.9–67.5% compared with higher keV levels (100–120 keV). Sharper kernels (Bv72) decreased the CNR by 56.9% but increased the MA by 54.3% compared with smoother kernels (Bv40). QIR4 enhanced the SNR by 36.5%, the CNR by 37.6%, and the MA by 0.94% compared with QIR3.

The interaction plots (Figure 2) revealed critical nonlinear relationships: at 40 keV, MA values peaked across all Bv levels, with Bv72 and Bv64 yielding the highest values, followed by Bv48, Bv56, and Bv40. Specifically, Bv48 produced an almost 25% higher MA than Bv40 at 40 keV. As keV increased to 120 keV, the MA values declined sharply and converged across all Bv levels, eliminating the initial difference between Bv48 and Bv40. At 40 keV, QIR4 exhibited a significantly higher SNR than QIR3 at most Bv levels, with Bv40 showing the highest SNR. This advantage diminished at 120 keV, with the SNR values decreasing across all conditions. Bv40 maintained the highest SNR across the entire keV energy. The CNR was maximized at the combination of 40 keV/Bv40/QIR4 (Figure 2C). At 40 keV, QIR4 significantly boosted the CNR across all Bv levels compared to QIR3, with Bv56 showing the largest gains. As keV increased, the CNR decreased monotonically for all Bv/QIR combinations, with the relative benefit of QIR4 diminishing at higher keV.

Figure 2 Interaction effects between keV and Bv for the MA, SNR, and CNR. (A) The interaction effect between keV and Bv for the MA; (B) the interaction effect between keV and Bv for the SNR; (C) the interaction effect between keV and Bv for the CNR. Bv, body vascular convolution; CNR, contrast-to-noise ratio; MA, mean arterial signal intensity attenuation; SNR, signal-to-noise ratio.

The comprehensive heatmap visualization (Figure 3) revealed distinct responses of the MA, SNR, and CNR to variations in keV, Bv, and QIR settings. The MA exhibited a strong negative correlation with keV, peaking at 40 keV and declining sharply to minimum values at 120 keV. Higher Bv levels (64 and 72) marginally increased the MA at 40 keV, while QIR settings (3 vs. 4) exerted negligible effects on the MA across all conditions. Conversely, the SNR and CNR remained uniformly low and stable across the entire parameter space (40–120 keV, Bv40–72, and QIR3/4).

Figure 3 Heatmap of parameter effects on the image quality metrics (the MA, SNR, and CNR). Bv, body vascular convolution; CNR, contrast-to-noise ratio; MA, mean arterial signal intensity attenuation; SNR, signal-to-noise ratio.

The composite scoring system identified five optimal configurations (Figure 4, Figure S2, and Table 3). Figure 5 shows the top five parameter configurations of PCD-CT images from a 71-year-old male patient with hypertension.

Figure 4 Performance comparison of top five parameter configurations. Bv, body vascular convolution; QIR, quantum iterative reconstruction.

Figure 5 Image examples for top five parameter configurations. A 71-year-old male patient with hypertension. Top row (A-E): 40-keV VMIs of the right OA with various kernels and QIR strength level combinations. Bottom row (F-J): 40-keV VMIs of the left OA with various kernels and QIR strength level combinations. The VMIs at 40 keV with Bv56 and QIR4 can better visualize the OA and branches, including the lacrimal artery (black arrow), and posterior ciliary artery (white arrow). In addition, the stenosis of the right OA (red star) was more clearly visualized. OA, ophthalmic artery; QIR, quantum iterative reconstruction; VMI, virtual monoenergetic imaging.

Subjective analysis

Both readers demonstrated highly consistent evaluation patterns across all image quality features. The Bv56/QIR4 protocol consistently achieved perfect median scores (5/5) with an IQR of zero for overall image quality, vessel sharpness, and diagnostic assess-ability. Intermediate reconstruction kernel protocols (Bv48) demonstrated moderate performance (median, 3–4; range, 3–4), while low reconstruction kernel protocols (Bv40) yielded substantially inferior results (median, 1–2; range, 1–2). Friedman tests confirmed statistically significant differences among protocols across all quality domains (P<0.001).

Post-hoc analysis (Dunn’s test with Bonferroni adjustment) revealed three key patterns: First, the Bv56 protocols significantly outperformed all lower reconstruction kernel combinations across both readers and all quality features (adjusted P<0.0001). Second, the Bv48 protocols demonstrated superior performance to the Bv40 protocols (adjusted P<0.0001). Third, no significant differences emerged between QIR3 and QIR4 at equivalent reconstruction kernel levels (adjusted P>0.05), though QIR3 showed marginally higher sharpness scores at Bv48 (Reader 1: median 4 vs. 3). The performance advantage of Bv56/QIR4 was most pronounced for diagnostic assess-ability.

Inter-reader concordance was exceptionally high, with both readers identifying identical protocol rankings (Bv56 > Bv48 > Bv40). Based on the maximum median scores and statistical significance, 40 keV/Bv56/QIR4 was established as the optimal reconstruction protocol, demonstrating 25–30% higher median scores than Bv48 alternatives, while maintaining perfect consistency across all quality features (Table 4 and Figure 6).

Figure 6 Subjective analysis scores for overall image quality, vessel sharpness, and diagnostic assess-ability of the OA in two readers using top five parameter configurations. The box indicates the median score, and the error bar shows the score range. * indicates significant differences with post-hoc pairwise comparison. *, P<0.05; **, P<0.01; ***, P<0.001. OA, ophthalmic artery.

Discussion

This study evaluated PCD-CT parameter optimization for OA imaging, revealing critical interactions between VMI, reconstruction kernels, and QIR levels. The study demonstrated that 40-keV VMI reconstruction achieved optimal results in both objective image quality metrics and subjective evaluation. Sartoretti et al. demonstrated that 40-keV reconstructions with QIR-4 maximize objective image quality in coronary CT (26). Similarly, our study confirmed that 40-keV VMIs optimize quantitative metrics, significantly enhancing edge sharpness, the CNR, and the SNR.

This sharpness improvement is partially attributable to inherent low-keV physics: the attenuation difference between iodine-enhanced blood and adjacent tissues amplifies at lower energies, steepening the attenuation gradient across vessel walls (27,28). However, this attenuation benefit comes at the expense of increased image noise, as reflected by the observed vessel attenuation reductions at lower keV levels. The interaction analysis further revealed that noise penalties were not uniform but were modulated by reconstruction kernel and QIR strength. These results emphasize that low-keV VMI should not be interpreted in isolation; rather, its diagnostic value depends on synergistic optimization with advanced reconstruction techniques.

Among all the parameters studied, reconstruction kernel sharpness emerged as the most influential factor for both objective and subjective image quality. The consistent superiority of the Bv56 protocols across readers and quality domains underscores the clinical relevance of high-frequency information preservation in OA assessment. While sharper kernels yielded higher MA values, they failed to provide sufficient vessel delineation and diagnostic confidence, as evidenced by markedly lower subjective scores (29,30). These observations reinforce the notion that vessel sharpness and structural clarity, rather than attenuation alone, are the primary determinants of diagnostic utility in UHR vascular imaging (31).

From a technical perspective, the OA typically approaches the spatial resolution limit of conventional CTA, rendering it highly susceptible to partial volume effects (32). Smoother reconstruction kernels, although capable of preserving higher attenuation values through spatial averaging, inherently blur high-frequency information and attenuate vessel boundaries. In our study, this phenomenon translated into inferior vessel sharpness and diagnostic assess-ability scores, despite relatively higher MA values. Conversely, sharper kernels (particularly Bv56) enhanced high-frequency signal preservation, allowing more accurate delineation of vessel margins and luminal continuity, which are critical for confident interpretation in clinical practice (33-35).

Our results align with and extend previous observations from high-resolution and ultra-high-resolution CT studies, which have demonstrated that sharper reconstruction kernels improve the visualization of fine vascular and osseous structures at the expense of increased noise. However, PCD-CT introduces a critical paradigm shift by decoupling spatial resolution from excessive noise amplification through improved detector efficiency and advanced iterative reconstruction (35,36). In this context, the combination of a sharp kernel with QIR enabled the simultaneous enhancement of vessel sharpness and acceptable noise levels, a balance that is difficult to achieve with conventional EID-CT systems (37). However, no direct comparison was made with conventional EID-CT. Technical studies have consistently highlighted the inherent advantages of PCD-CT over EID-CT, including higher spatial resolution, a better CNR, superior spectral separation, and reduced electronic noise (38,39). These properties theoretically benefit the imaging of small structures such as the OA.

Our findings suggest QIR4 reconstruction offers significant benefits for OA visualization in low-keV PCD-CT imaging. Quantitative analysis indicated that QIR4 typically increased the SNR by approximately 36% (30.47 vs. 22.33) and the CNR by approximately 38% CNR (24.46 vs. 17.78) compared to QIR3, with observed noise reduction of approximately 28% across VMI levels. QIR4 contributed to improved image interpretability in the challenging 40–60-keV range, reducing the incidence of noise-related non-diagnostic ratings (40). QIR4 may help strike a favorable balance between noise reduction and structural fidelity for small-vessel assessment (41,42). While not universally required, its application appears particularly advantageous when implementing low-energy OA protocols, potentially enabling diagnostic confidence at energy levels that might otherwise prove challenging for clinical interpretation.

Our findings suggest that the 40-keV/Bv56/QIR4 protocol may offer a balanced approach for OA visualization, helping reconcile differences between quantitative metrics and clinical usability. These collective adjustments may enable the diagnostic assessment of small-vessel structures that might otherwise be challenging to evaluate consistently with conventional approaches.

This study had several limitations. First, the small cohort size may limit the detection of subtle parameter effects. Second, all examinations were performed on a single PCD-CT system (Siemens NAEOTOM Alpha), but reconstruction parameter performance may vary between scanner platforms, which may affect generalizability. Third, the comparative analysis was limited to PCD-CT reconstruction techniques; the lack of direct comparisons with conventional EID-CT limits a specific understanding of the relative benefits of PCD-CT. Fourth, our study used a fixed image quality reference level (CARE Dose4D level 145) for acquisition, which determines the radiation dose and consequently the baseline quantum noise level in the projection data. Future studies systematically evaluating the impact of different dose levels (image quality reference settings) on these reconstruction parameters for PCD-CT OA imaging would be valuable to fully characterize these dependencies and identify optimal protocols across a range of acceptable radiation exposures. Finally, protocol optimization focused only on normal OA anatomy. Future studies should evaluate parameter performance in pathological conditions (e.g., arterial occlusion, stenosis, or aneurysm) where parameter performance is clinically relevant and optimal visualization is particularly important.

Conclusions

This preliminary investigation suggests that PCD-CT, using a reconstructed energy level of 40 keV combined with the reconstruction kernel of Bv56 and QIR level 4 reconstruction, may offer a balanced approach for visualizing the OA. The protocol appears to provide sufficient vascular contrast while managing image noise, potentially aiding in the delineation of this small-caliber vessel. Ultimately, the continued refinement of PCD-CT protocols holds promise for orbital small-vessel assessment.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2026-1-0247/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2026-1-0247/dss

Funding: This study was supported by the Key Laboratory of Functional Molecular Imaging of Tumor and Interventional Diagnosis and Treatment of Shaoxing City (No. 2020ZDSYS01) and Shaoxing Medicine and Health Science and Technology Project (No. 2024SKY016).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2026-1-0247/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Academic Ethics Committee of Shaoxing People’s Hospital, which waived the requirement to obtain patient approval or written informed consent for the review of medical records or images.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Hayreh SS. Orbital vascular anatomy. Eye (Lond) 2006;20:1130-44. [Crossref] [PubMed]
2. Tufek M, Nalcacioglu P, Capraz M, Varol K, Kaya AT, Aydın N, Kara C. The impact of obesity on ocular hemodynamics and choroidal thickness. Ther Adv Ophthalmol 2023;15:25158414231180985. [Crossref] [PubMed]
3. Wu LT, Wang JL, Wang YL. Ophthalmic artery changes in type 2 diabetes with and without acute coronary syndrome. J Transl Med 2022;20:512. [Crossref] [PubMed]
4. Sirkeci O, Erkus Sirkeci E, Ergün Süzer N, Tatlıparmak AC, Küçükçiloğlu Y, Ulaş T. Can orbital artery resistive indexes detect complications of diabetes in the prediabetic phase? Wien Klin Wochenschr 2023;135:343-8. [Crossref] [PubMed]
5. Perrini P, Cardia A, Fraser K, Lanzino G. A microsurgical study of the anatomy and course of the ophthalmic artery and its possibly dangerous anastomoses. J Neurosurg 2007;106:142-50. [Crossref] [PubMed]
6. Zhang T, Fan S, He W, Zhang T, Wang Y. Ophthalmic artery visualization and morphometry by computed tomography angiography. Graefes Arch Clin Exp Ophthalmol 2015;253:627-31. [Crossref] [PubMed]
7. Lang J, Kageyama I. The ophthalmic artery and its branches, measurements and clinical importance. Surg Radiol Anat 1990;12:83-90. [Crossref] [PubMed]
8. Raz E, Shapiro M, Shepherd TM, Nossek E, Yaghi S, Gold DM, Ishida K, Rucker JC, Belinsky I, Kim E, Mac Grory B, Mir O, Hagiwara M, Agarwal S, Young MG, Galetta SL, Nelson PK. Central Retinal Artery Visualization with Cone-Beam CT Angiography. Radiology 2022;302:419-24. [Crossref] [PubMed]
9. Akdemir Aktaş H, Mine Ergun K, Tatar İ, Arat A, Mutlu Hayran K. Investigation into the ophthalmic artery and its branches by superselective angiography. Interv Neuroradiol 2022;28:737-45. [Crossref] [PubMed]
10. DI CHIRO G. Ophthalmic arteriography. Radiology 1961;77:948-57. [Crossref] [PubMed]
11. Farnsworth PJ, Campeau NG, Diehn FE, Yu L, Leng S, Zhou Z, Fletcher JG, McCollough CH. High-resolution computed tomography angiography of the orbit using a photon-counting computed tomography scanner. Interv Neuroradiol 2025;31:539-54. [Crossref] [PubMed]
12. van der Bie J, van Straten M, Booij R, Bos D, Dijkshoorn ML, Hirsch A, Sharma SP, Oei EHG, Budde RPJ. Photon-counting CT: Review of initial clinical results. Eur J Radiol 2023;163:110829. [Crossref] [PubMed]
13. Willemink MJ, Persson M, Pourmorteza A, Pelc NJ, Fleischmann D. Photon-counting CT: Technical Principles and Clinical Prospects. Radiology 2018;289:293-312. [Crossref] [PubMed]
14. Wu Y, Ye Z, Chen J, Deng L, Song B. Photon Counting CT: Technical Principles, Clinical Applications, and Future Prospects. Acad Radiol 2023;30:2362-82. [Crossref] [PubMed]
15. He N, Zhang Y, Li Z, Xu Z, Lyu H, Li J, Dong H, Zhu C, Haacke EM, Mossa-Basha M, Schmidt B, Jiang H, Yan F. Ultrahigh-Resolution Photon-Counting Detector CTA of the Head and Neck: Image Quality Assessment and Vascular Kernel Optimization. AJR Am J Roentgenol 2025;224:e2431763. [Crossref] [PubMed]
16. Booz C, Bucolo GM, D’Angelo T, Mazziotti S, Lanzafame LRM, Yel I, Alizadeh LS, Gruenewald LD, Koch V, Martin SS, Dimitrova M, Goekduman A, Vogl TJ, Kaatsch HL, Overhoff D, Waldeck S. Carotid artery assessment in dual-source photon-counting CT: impact of low-energy virtual monoenergetic imaging on image quality, vascular contrast and diagnostic assessability. Radiol Med 2024;129:1633-43. [Crossref] [PubMed]
17. Dillinger D, Overhoff D, Booz C, Kaatsch HL, Piechotka J, Hagen A, Froelich MF, Vogl TJ, Waldeck S. Impact of CT Photon-Counting Virtual Monoenergetic Imaging on Visualization of Abdominal Arterial Vessels. Diagnostics (Basel) 2023.
18. Dunning CAS, Rajendran K, Inoue A, Rajiah P, Weber N, Fletcher JG, McCollough CH, Leng S. Optimal Virtual Monoenergetic Photon Energy (keV) for Photon-Counting-Detector Computed Tomography Angiography. J Comput Assist Tomogr 2023;47:569-75. [Crossref] [PubMed]
19. Rajendran K, Petersilka M, Henning A, Shanblatt ER, Schmidt B, Flohr TG, Ferrero A, Baffour F, Diehn FE, Yu L, Rajiah P, Fletcher JG, Leng S, McCollough CH. First Clinical Photon-counting Detector CT System: Technical Evaluation. Radiology 2022;303:130-8. [Crossref] [PubMed]
20. Euler A, Higashigaito K, Mergen V, Sartoretti T, Zanini B, Schmidt B, Flohr TG, Ulzheimer S, Eberhard M, Alkadhi H. High-Pitch Photon-Counting Detector Computed Tomography Angiography of the Aorta: Intraindividual Comparison to Energy-Integrating Detector Computed Tomography at Equal Radiation Dose. Invest Radiol 2022;57:115-21. [Crossref] [PubMed]
21. Ren L, Duan X, Ahn R, Kay F, Daftaribesheli L, Zhou W, Guild J, Ananthakrishnan L. Photon-counting detector CT: technology overview and radiation dose reduction. Br J Radiol 2025;98:1788-801. [Crossref] [PubMed]
22. Leng S, Bruesewitz M, Tao S, Rajendran K, Halaweish AF, Campeau NG, Fletcher JG, McCollough CH. Photon-counting Detector CT: System Design and Clinical Applications of an Emerging Technology. Radiographics 2019;39:729-43. [Crossref] [PubMed]
23. Lasjaunias P, Berenstein A, Brugge KG. Clinical vascular anatomy and variations. Springer; 2001.
24. Tawfik AM, Kerl JM, Razek AA, Bauer RW, Nour-Eldin NE, Vogl TJ, Mack MG. Image quality and radiation dose of dual-energy CT of the head and neck compared with a standard 120-kVp acquisition. AJNR Am J Neuroradiol 2011;32:1994-9. [Crossref] [PubMed]
25. Waaijer A, Prokop M, Velthuis BK, Bakker CJ, de Kort GA, van Leeuwen MS. Circle of Willis at CT angiography: dose reduction and image quality--reducing tube voltage and increasing tube current settings. Radiology 2007;242:832-9. [Crossref] [PubMed]
26. Sartoretti T, McDermott M, Mergen V, Euler A, Schmidt B, Jost G, Wildberger JE, Alkadhi H. Photon-counting detector coronary CT angiography: impact of virtual monoenergetic imaging and iterative reconstruction on image quality. Br J Radiol 2023;96:20220466. [Crossref] [PubMed]
27. Qi K, Li L, Yuan D, Zhang Y, Zhang M, Zhang W, Gu Y, Wang Z, Lyu P, Gao J, Liu J. Optimized contrast enhancement and homogeneity in aortic CT angiography: bolus tracking with personalized post-trigger delay. Quant Imaging Med Surg 2025;15:709-20. [Crossref] [PubMed]
28. Schwartz FR, Bache S, Lee R, Maxfield CM, Fadell MF 2nd, Gaca AM, Samei E, Frush DP, Cao JY. Photon-counting CT yields superior abdominopelvic image quality at lower radiation and iodinated contrast doses. Pediatr Radiol 2025;55:1202-11. [Crossref] [PubMed]
29. Rippel K, Luitjens J, Habeeballah O, Scheurig-Muenkler C, Bette S, Braun F, Kroencke TJ, Schwarz F, Decker JA. Evaluation of ECG-Gated, High-Pitch Thoracoabdominal Angiographies With Dual-Source Photon-Counting Detector Computed Tomography. J Endovasc Ther 2025;32:2236-46. [Crossref] [PubMed]
30. Rippel K, Decker JA, Luitjens J, Habeeballah O, Bette S, Braun F, Kroencke TJ, Scheurig-Muenkler C. Virtual Monoenergetic Imaging of Thoracoabdominal Computed Tomography Angiography on Photon-Counting Detector Computertomography: Assessment of Image Quality and Leveraging Low-keV Series for Salvaging Suboptimal Contrast Acquisitions. Diagnostics (Basel) 2024;14:2843. [Crossref] [PubMed]
31. Farhadi F, Sahbaee P, Rajagopal JR, Nikpanah M, Saboury B, Gutjahr R, Biassou NM, Shah R, Flohr TG, Samei E, Pritchard WF, Malayeri AA, Bluemke DA, Jones EC. Virtual monoenergetic imaging in photon-counting CT of the head and neck. Clin Imaging 2023;102:109-15. [Crossref] [PubMed]
32. Soschynski M, Hagen F, Baumann S, Hagar MT, Weiss J, Krauss T, Schlett CL, von Zur Mühlen C, Bamberg F, Nikolaou K, Greulich S, Froelich MF, Riffel P, Overhoff D, Papavassiliu T, Schoenberg SO, Faby S, Ulzheimer S, Ayx I, Krumm P. High Temporal Resolution Dual-Source Photon-Counting CT for Coronary Artery Disease: Initial Multicenter Clinical Experience. J Clin Med 2022;11:6003. [Crossref] [PubMed]
33. Yang Y, Fink N, Emrich T, Graafen D, Richter R, Bockius S, Wolf EV, Laux G, Kavermann L, Müller L, Hell M, Halfmann MC. Optimization of Kernel Type and Sharpness Level Improves Objective and Subjective Image Quality for High-Pitch Photon Counting Coronary CT Angiography. Diagnostics (Basel) 2023;13:1937. [Crossref] [PubMed]
34. Ucar FA, Frenzel M, Kronfeld A, Altmann S, Sanner AP, Mercado MAA, Uphaus T, Brockmann MA, Othman AE. Improvement of Neurovascular Imaging Using Ultra-High-Resolution Computed Tomography Angiography. Clin Neuroradiol 2024;34:189-99. [Crossref] [PubMed]
35. Tóth A, Chetta JA, Yazdani M, Matheus MG, O’Doherty J, Tipnis SV, Spampinato MV. Neurovascular Imaging with Ultra-High-Resolution Photon-Counting CT: Preliminary Findings on Image-Quality Evaluation. AJNR Am J Neuroradiol 2024;45:1450-7. [Crossref] [PubMed]
36. Masturzo L, Barca P, De Masi L, Marfisi D, Traino A, Cademartiri F, Giannelli M. Voxelwise characterization of noise for a clinical photon-counting CT scanner with a model-based iterative reconstruction algorithm. Eur Radiol Exp 2025;9:2. [Crossref] [PubMed]
37. Gruschwitz P, Hartung V, Kleefeldt F, Ergün S, Lichthardt S, Huflage H, Hendel R, Kunz AS, Pannenbecker P, Kuhl PJ, Augustin AM, Bley TA, Petritsch B, Grunz JP. Standardized assessment of vascular reconstruction kernels in photon-counting CT angiographies of the leg using a continuous extracorporeal perfusion model. Sci Rep 2023;13:12109. [Crossref] [PubMed]
38. Barca P, Masturzo L, De Masi L, Traino A, Cademartiri F, Giannelli M. Characterization of the Physical Image Quality of a Clinical Photon-Counting Computed Tomography Scanner Across Multiple Acquisition and Reconstruction Settings. Appl Sci 2026;16:1322. [Crossref]
39. Bhattarai M, Bache S, Abadi E, Samei E. A systematic task-based image quality assessment of photon-counting and energy integrating CT as a function of reconstruction kernel and phantom size. Med Phys 2024;51:1047-60. [Crossref] [PubMed]
40. van der Bie J, Bos D, Dijkshoorn ML, Booij R, Budde RPJ, van Straten M. Thin slice photon-counting CT coronary angiography compared to conventional CT: Objective image quality and clinical radiation dose assessment. Med Phys 2024;51:2924-32. [Crossref] [PubMed]
41. Vattay B, Boussoussou M, Vecsey-Nagy M, Kolossváry M, Juhász D, Kerkovits N, Balogh H, Nagy N, Vértes M, Kiss M, Kubovje A, Merkely B, Maurovich Horvat P, Szilveszter B. Qualitative and quantitative image quality of coronary CT angiography using photon-counting computed tomography: Standard and Ultra-high resolution protocols. Eur J Radiol 2024;175:111426. [Crossref] [PubMed]
42. Michael AE, Schoenbeck D, Woeltjen MM, Boriesosdick J, Kroeger JR, Moenninghoff C, Horstmeier S, Niehoff JH, Kabbasch C, Goertz L, Borggrefe J. Nonenhanced Photon Counting CT of the Head: Impact of the keV Level, Iterative Reconstruction and Calvaria on Image Quality in Monoenergetic Images. Clin Neuroradiol 2024;34:75-83. [Crossref] [PubMed]