Abstract
In this single-center, retrospective cohort study, we compared lower extremity photon-counting detector computed tomography (PCD-CT; n = 43) and conventional energy-integrating detector CT angiograms (EID-CTA; n = 46) with traditional on-table angiograms to assess whether there was a difference in the ability to characterize the target tibial artery (TA) for potential lower extremity bypass between imaging modalities. In PCD-CT angiography vs angiography, a different TA was identified in two patients (4.6%), whereas in EID-CTA vs angiography, a different TA was identified in 10 patients (21.7%; P = .03). PCD-CT angiography may allow for more accurate identification of the TA for lower extremity bypass than EID-CTA.
Keywords: Peripheral arterial disease, CT, Angiogram, Photon counting, Bypass
Evaluation of the infrapopliteal arteries can be challenging using traditional energy-integrating detector computed tomography angiography (EID-CTA) in patients with peripheral arterial disease (PAD). This is due to the small size and calcification of the tibial vessels, which create significant calcium blooming artifacts on EID-CTA, making it challenging to distinguish calcium and contrast. Additionally, it is difficult to obtain strong contrast opacification of tibial vessels on EID-CTA due to the presence of occlusive disease, the high Hounsfield units of calcium resulting in saturated pixels, and venous contamination in patients with PAD or critical limb-threatening ischemia.1 For these reasons, EID-CTA often cannot provide sufficient evaluation of the tibial vessels for lower extremity bypass planning, and on-table angiography remains the gold standard.
Photon-counting detector CTA (PCD-CTA) is an exciting new technology that has potential advantages over EID-CTA for evaluation of the infrapopliteal arteries.2 PCD-CTA differs from conventional EID-CTA in several ways. In conventional EID-CTA, EIDs convert incoming x-rays into visible light as they pass through a physical reflective septum, and then into an electrical signal (Fig 1, A).1,3 In doing so, EID-CTA compiles the total deposited energy of the incoming x-ray, rather than distinguishing individual photon energies.1,3 PCD-CTA, in contrast, uses energy-resolving detectors that directly convert each detected photon into an electrical signal, which is counted individually (Fig 1, B). This individual conversion of photons into electrical signals enables improved spatial resolution.1,3 This improved spatial resolution reduces the calcium blooming artifact, which in turn enables better visualization of small, calcified arteries such as the tibial vessels on PCD-CTA compared with EID-CTA.1 PCD-CTA also provides greater image contrast for iodine-containing materials, such as blood vessels, again due to the equal weighting of lower energy photons, compared with traditional EID-CTA. PCD-CTA also has lower electronic noise due to thresholding of photons at a lower energy level of noise. Due to these features, PCD-CTA can generate images of comparable noise to EID-CTA with a lower radiation dose or images with lower noise at the same radiation dose.4
Fig 1.
Schematic demonstrating the key technical aspects and differences between photon counting detector computed tomography angiography (PCD-CTA) and standard energy integrating detector CTA (EID-CTA).
Given this background, we hypothesized that PCD-CTA may enable better identification of tibial target artery (TA) for lower extremity bypass planning compared with EID-CTA. In this retrospective, single-center study, we aimed to compare lower extremity PCD-CTA and EID-CTA with traditional, gold standard on-table angiography for characterizing lower extremity runoff. We also sought to assess whether there is a difference in the ability to characterize tibial TAs for lower extremity bypass between PCD-CTA and EID-CTA.
Methods
This was a retrospective, single-institution study conducted between January 1, 2023, and January 1, 2025. Approval was obtained from our local institutional review board. These inclusion dates were selected based on the timeline that PCD-CTA technology was consistently available and validated at our institution, as described by Yalon et al,1 for the evaluation of the lower extremity vasculature in a prospective study. Patients who underwent bilateral lower extremity runoff PCD-CTA or EID-CTA and an on-table angiography during the study period were included. Patients who had incomplete imaging (such as an angiogram that did not include the foot) or interval limb loss were excluded. During the study period at our institution, whether a patient had a PCD-CTA or an EID-CTA performed was based on whether it was specified by the ordering physician, or protocolized that way by the radiologist, at their discretion.
All images were reviewed by two independent physicians: one vascular surgery attending physician and one senior-level vascular surgery resident. All images were reviewed by both reviewers independently. At the time of image evaluation, the physicians were blinded to the patient's clinical details and the other image types for that patient. For example, the physicians did not look at the patient's CT scans and on-table angiograms concurrently. The reviewers were also blinded to each other's reviews until the time of data analysis. Each reviewer generated an independent interpretation of all images and did not access or rely on prior radiological reports for any of the images.
For each imaging modality, the reviewers determined what the TA pathway for lower extremity bypass would be. The TA pathway for each imaging modality (PCD-CTA, EID-CTA, or on-table angiography) was determined separately by each reviewer. The Global Anatomic Staging System (GLASS) classification (as described by Conte et al5) was then assigned to each arterial segment, based on the TA pathway. This included the femoral-popliteal segment, the infrapopliteal/tibial segment, and the inframalleolar/pedal segment. The GLASS criteria for each infrainguinal arterial segment are specified in Table I, as described by Conte et al.5 The GLASS system was developed to define the preferred target artery pathway for a lower extremity bypass that can then be used to estimate limb-based patency based on the overall GLASS score.5
Table I.
Global Anatomic Staging System (GLASS) classification criteria for each arterial segment at each level
| FP GLASS grade | IP GLASS grade | Inframalleolar/pedal descriptor | ||
|---|---|---|---|---|
| FP grade 0 | Mild or no significant (<50%) disease | IP grade 0 | Mild or no significant disease | P0: TA crosses ankle into foot with intact pedal arch |
| FP grade 1 | Total length SFA disease less than one-third of the total vessel length (<10 cm) Single focal total occlusion (<5 cm) as long as not flush occlusion Popliteal artery with mild or no significant disease |
IP grade 1 | Focal stenosis of the tibial artery <3 cm | P1: TA crosses the ankle into the foot, absent or severely diseased pedal arch |
| FP grade 2 | Total length SFA disease one-third to two-thirds of the total vessel length (10-20 cm) May include occlusion totaling less than one-third of the total vessel length (10 cm), but not flush occlusion Focal popliteal artery stenosis <2 cm, not involving trifurcation |
IP grade 2 | Stenosis involving one-third of the total vessel length Focal occlusion (<3 cm) not including the tibioperoneal trunk or tibial vessel origin |
P2: No target artery crossing the ankle into the foot |
| FP grade 3 | Total length SFA disease greater than two-thirds of the total vessel length (>20 cm) length May include any flush occlusion <20 cm or nonflush occlusion 10-20 cm long Short popliteal stenosis 2-5 cm, not involving trifurcation |
IP grade 3 | Disease up to two-thirds of the vessel length Occlusion up to one-third of the total vessel length (may include tibial vessel origin but not the tibioperoneal trunk) |
|
| FP grade 4 | Total length SFA occlusion >20 cm Popliteal disease >5 cm or extending into trifurcation Any popliteal total occlusion |
IP grade 4 | Diffuse stenosis of more than two-thirds of the total vessel length Occlusion of more than one-third of the vessel length Any occlusion of tibioperoneal trunk |
FP, Femoral-popliteal; IP, infrapopliteal; SFA, superficial femoral artery; TA, target tibial artery.
The GLASS classification criteria for each arterial segment at each level are presented as described by Conte et al.5 The reviewers used these criteria specifically for calculating the GLASS score for each arterial segment.
GLASS classification and TA pathway identification for the two reviewers were then analyzed. Which tibial TA was identified as well as the GLASS classification for each infrainguinal arterial segment were compared for each reviewer. These were reviewed and compared separately for each imaging modality on each patient. We then compared the results for the PCD-CTAs and EID-CTAs with the on-table angiograms for each patient, and we also compared the results for the PCD-CTAs and EID-CTAs with each other for each patient. We then compared the prevalence (described as number [%]) of divergence between reviewers on the GLASS score for each arterial segment, with divergence being defined as the difference in the GLASS score for a single arterial segment that would change the total GLASS score for the patient. For example, if the difference in GLASS score between the reviewers for a single arterial segment would increase or decrease the overall GLASS score for the patient, that was considered a divergence. The χ2 test and Fisher's exact t-test were used for comparison of categorical variables, as appropriate. The Wilcoxon rank-sum test or Student t-test were used as appropriate for continuous variables. Data were analyzed using GraphPad Prism (GraphPad Software). Results are expressed as medians with quartiles [Q1, Q3], unless otherwise specified.
Results
During the study period, there were 43 patients identified who underwent PCD-CTA and on-table angiography; 1 patient was excluded due to incomplete imaging, and 42 patients were included in the analysis. There were 46 patients identified who had undergone EID-CTA and on-table angiography during the study period; 4 were excluded (1 due to interval below-knee amputation, 3 due to incomplete imaging), and 42 patients were included in the analysis. The cohort was 77% male, 93% self-identified as White, and the median age at intervention was 71 years [61, 76 years]. As described in Table II, there was no significant difference in the rates of baseline comorbidities between the EID-CTA and PCD-CTA patient groups. Example images from patients included in this study can be found in Fig 2, which includes a side-by-side comparison of images from EID-CTA and PCD-CTA of the tibial vessels. This figure highlights the sharper image resolution of PCD-CTA compared with EID-CTA, and the ability of PCD-CTA to demonstrate the architecture of the calcium in the tibial vessel wall (Figs 2, A-C).
Table II.
Baseline clinical characteristics of the patient cohort
| EID-CTA (n = 42) | PCD-CTA (n = 42) | P value | |
|---|---|---|---|
| Diabetes | 25 | 21 | .08 |
| Hypertension | 47 | 44 | .10 |
| Chronic kidney diseasea | 10 | 12 | .30 |
| Coronary artery disease | 20 | 19 | .10 |
| Hyperlipidemia | 40 | 34 | .09 |
| Active smokers | 11 | 14 | .20 |
| Former smokers | 30 | 25 | .07 |
Values are percent. P values of <.05 were considered statistically significant.
There were no patients on hemodialysis in this patient cohort.
Fig 2.
(A–C) Side-by-side comparison of photon-counting detector computed tomography angiography (PCD-CTA) and energy-integrating detector CTA (EID-CTA) imaging of the infrapopliteal arteries (each pair from the same patient) including (A) the posterior tibial artery at the foot, (B) the peroneal and anterior tibial arteries in the distal calf, and (C) all three tibial vessels in the proximal calf. These images are from patients in this study.
Among the cohort analyzed (n = 84), 35 patients (42%) presented with tissue loss, 18 (21%) presented with rest pain, and 21 (25%) presented with lifestyle-limiting short distance claudication. Fourteen patients (17%) were undergoing evaluation for changes in surveillance imaging or injury. Fourteen patients (17%) had undergone prior lower extremity vascular intervention on the ipsilateral leg of interest. Seven of these were endovascular infrainguinal interventions and seven of these were femoral endarterectomies with or without retrograde iliac intervention.
Each reviewer separately determined what the target or primary runoff TA would be for lower extremity bypass on each imaging modality, for each patient. The two reviewers independently agreed on what the target or primary runoff TA would be for each patient, on each imaging modality. The reviewers identified a different target or primary runoff TA on EID-CTA compared with the angiogram more frequently than with PCD-CTA compared with the angiogram (2 [4.7%] vs 10 [24.0%]; P = .03) on the same patient. For example, on the same patient, the reviewer might identify the anterior tibial artery as the TA on the CT scan but the peroneal artery as the TA on the angiogram more frequently on the EID-CTA than on the PCD-CTA. Fig 3 demonstrates an example of this, where the posterior tibial artery appeared to be the TA on EID-CTA (Fig 3, A), but then the anterior tibial artery was found to be the better TA with on-table angiography (Fig 3, B), as the posterior tibial artery was found to taper off just below the ankle.
Fig 3.
(A, B) Side-by-side comparison of energy-integrating detector computed tomography angiography (EID-CTA) (A) and an on-table angiography (B) of the distal tibial vessels at the ankle and proximal foot for the same patient. Based on these images, the target artery pathway was determined to be the posterior tibial artery on the EID-CTA (A, arrow) vs the anterior tibial artery on the gold standard angiography (B), as the posterior tibial artery was found to taper off at the ankle on the on-table angiogram.
The GLASS classification scores for each infrainguinal arterial segment were also compared on each imaging modality for each patient. There was no significant difference between the mean GLASS score on the PCD-CTAs vs on-table angiograms for the femoral-popliteal segment (2.3 ± 1.6 vs 2.2 ± 1.6; P = .20) or for the tibial segment (1.8 ± 1.3 vs 1.5 ± 1.4; P = .30). There was no significant difference between the mean GLASS score on EID-CTAs vs on-table angiograms for the femoral-popliteal segment (1.8 ± 1.4 vs 1.9 ± 1.5; P = .30). However, the mean GLASS score assigned for the tibial segment was significantly higher on the EID-CTA than on the on-table angiogram (1.9 ± 1.5 vs 1.2 ± 1.4; P ≤ .01).
There were no significant differences in the rates of divergence in GLASS scores between the reviewers for the femoral-popliteal or tibial segments on PCD-CTA vs angiography or EID-CTA vs angiography (Table III). When PCD-CTA and EID-CTA were compared with each other, there was no significant difference in the rates of GLASS score divergence between the femoral-popliteal segment, but there was significantly greater divergence in GLASS scores between reviewers for the tibial and inframalleolar segments (Table III).
Table III.
Comparison of the prevalence of Global Anatomic Staging System (GLASS) score divergence
| Rate of GLASS score divergence between reviewers | PCD-CTA vs angiogram (n = 42) | EID-CTA vs angiogram (n= 42) | P value |
|---|---|---|---|
| Femoral-popliteal | 7 (17) | 7 (17) | 1.0 |
| Tibial | 14 (33) | 19 (45) | .18 |
| Rate of GLASS score divergence between reviewers | PCD-CTA (n = 42) | EID-CTA (n = 42) | |
|---|---|---|---|
| Femoral-popliteal | 4 (9.5) | 4 (9.5) | 1.0 |
| Tibial | 6 (14) | 15 (35) | .04 |
| Inframalleolar | 0 (0.0) | 8 (19) | <.01 |
EID-CTA, Energy-integrating detector computed tomography angiography; PCD-CTA, photon-counting detector computed tomography angiography.
Values are number (%).
A P value of <.05 was considered significant.
Comparisons are between the two reviewers for each arterial segment, comparing PCD-CTA, EID-CTA and gold standard on-table angiography. We compared the prevalence of divergence between reviewers on the GLASS score for each arterial segment, with divergence being defined as a difference in the GLASS score for a single arterial segment that would change the total GLASS score for the patient. For example, if the difference in GLASS score between the reviewers for a single arterial segment would increase or decrease the overall GLASS score for the patient, that was considered a divergence.
Inframalleolar comparison not included for PCD-CTA and EID-CTA vs angiography due to incomplete foot shots in some angiograms.
Discussion
PCD-CT offers a noninvasive option to obtain high-quality imaging of small, calcified vessels, such as the tibial arteries. Although on-table angiography remains the gold standard for identification of the TA for lower extremity bypass planning, PCD-CTA can provide excellent imaging of the tibial arteries without invasive imaging, which can be a helpful adjunct to on-table angiography. PCD-CTA can also provide tibial runoff information quickly in an urgent setting when preoperative on-table angiography is not available. PCD-CTA shows the architecture of the calcium within the tibial arteries, as can be seen in Fig 2. Given that the TA is often diseased in patients with PAD undergoing lower extremity bypass, PCD-CTA can allow one to assess the distribution of calcium within the vessel, which can be helpful in planning exactly where to put the distal anastomosis or anticipating the need for focal endarterectomy. In this way, PCD-CTA can provide greater detail on tibial calcium distribution compared with on-table angiography.
The high-quality high-resolution images generated by PCD-CTA have been beneficial in cardiovascular imaging, thoracic imaging, and pediatric imaging.4 PCD-CTA may also enable higher resolution images without a major radiation dose increase in patients with a high body mass index.4 PCD-CTA has been shown to have benefit and increasing application in coronary imaging. For coronary artery calcium scoring scans, PCD-CTAs have been shown to have radiation dose reductions between 25% and 50%, depending on the protocol used, compared with EID-CTA or other high-quality CT imaging protocols.6,7 PCD-CTA has also enabled the acquisition of ultra-high-quality images in the coronary arteries, which allows for the assessment of luminal stenosis associated with dense calcified plaques and stents at a level akin to an invasive coronary catheterization.8, 9, 10 Like the coronary arteries, in patients with PAD, the tibial vessels are small and calcified. The experience in coronary imaging demonstrates the level of potential of PCD-CTA as a tool in the workup of PAD.
Early studies examining PCD-CTA and EID-CTA for imaging of the infrapopliteal arteries remain small but promising. In a prospective study of 32 patients who underwent clinically indicated EID-CTA followed by investigational PCD-CTA that same day (at a lower contrast dose), two radiologists were independently able to identify significantly more fibular perforators on PCD-CTA than EID-CTA.1 They were also able to more easily distinguish between stenotic areas and total occlusion of the infrapopliteal vessels and were more easily able to identify diseased infrapopliteal vessels on PCD-CTA compared with EID-CTA. In that study, the radiologists more frequently overestimated the degree of infrapopliteal disease on the EID-CTA compared with the PCD-CTA, as they identified more total occlusions on EID-CTA that were read as stenosis on the PCD-CTA scans.1 Another study comparing PCD-CTA and EID-CTA for imaging of stented peripheral arteries showed that PCD-CTA enables high-quality imaging with reduced noise and greater radiation dose efficacy in stented lower extremity vessels, which like the tibial vessels can be challenging to evaluate with standard EID-CT.11 Further work in this area is needed to evaluate the full potential of PCD-CT technology in PAD management.
In this study, the two reviewers differed more frequently on identification of the target or primary runoff TA with EID-CTA compared with PCD-CTA vs on-table angiography. In 24% of cases in this series, the reviewers identified a different TA on the EID-CTA than on the on-table angiogram. With PCD-CTA, there was discrepancy in TA identification in only 4.7% of cases. This finding demonstrates that PCD-CTA allows for more reliable identification of the TA for lower extremity bypass than EID-CTA when an on-table angiogram is used as the gold standard.
The reviewers assigned a GLASS classification score to each infrainguinal arterial segment. We used variability in the GLASS scores between imaging modalities and variance between the reviewers in GLASS scores as surrogate markers for the reliable characterization of lower extremity runoff. There was no significant difference in the mean GLASS scores assigned by the reviewers for the femoral-popliteal or tibial segments when the PCD-CTAs and on-table angiograms were compared. However, when EID-CTA was compared with an on-table angiogram, the tibial GLASS score was significantly higher on the EID-CTA than the on-table angiogram. This finding indicates that the reviewers tended to overestimate the degree of tibial disease on the EID-CTAs compared with the on-table angiograms, whereas PCD-CTA gave a more concordant picture of the tibial vessels with an on-table angiogram. This result is likely because of the significant calcium blooming artifact present on the EID-CTA, which makes the tibial vessels difficult to assess. When evaluated by EID-CTA, the tibial vessels may appear to be completely calcified or of unclear patency, whereas PCD-CTA allows for improved visualization due to reduced calcium blooming artifact. Similarly, when we examined variance in GLASS classification scores between the reviewers for each infrainguinal arterial segment, we found that there was significantly greater variability between the reviewers in the tibial and inframalleolar GLASS scores between the PCD-CTAs and EID-CTAs. This result again demonstrates that the reviewers were able to more consistently and reliably characterize the infrapopliteal arteries on PCD-CTA than on EID-CTA.
This study has several limitations. This is a single-center, retrospective study that therefore has the limits of any retrospective series. Additionally, this is a small series with a total of 84 patients; therefore, the number of pathologies included in this series is limited, and it may not be applicable to all infrainguinal pathologies. In this study, scans were evaluated by two separate reviewers. Although the reviews of each arterial segment were compared, and all attempts were made to limit bias, there is always a potential for the individual interpretations of the reviewers to have been biased. Although the reviewers were blinded to most identifying, demographic, and clinical details for each patient when reviewing and analyzing the images, the reviewers did have access to each patient's medical record number to find their images in the electronic system. As such, recall bias for specific patients that the reviewers may have encountered before could exist. The reviewers in this series were both vascular surgeons, not radiologists, so that may have also introduced some degree of interpretation bias into this study.
Another limitation of this study is the generalizability and availability of PCD-CTA. Presently, PCD-CTA is commercially available on only one CT system (Siemens Healthcare Diagnostics). Existing EID-CT systems can be enhanced to perform PCD-CTA, but angiography is not currently available on other systems. PCD-CTA also has a higher cost than EID-CTA. This factor may vary between institutions, but at our institution this can range between one and one-half times to twice the cost of a conventional EID-CTA. This cost and availability issue limits the clinical applicability and availability of this technology at the current time.
Conclusions
PCD-CTA may allow for more accurate identification of the TA for lower extremity bypass than conventional EID-CTA, enabling more reliable characterization of infrapopliteal runoff in the absence of an on-table angiogram. This exciting new technology can be used as an adjunct to on-table angiography, which remains the gold standard, for the workup and treatment of patients with PAD who require lower extremity bypass.
Funding
None.
Disclosures
None.
From the Midwestern Vascular Surgical Society
Footnotes
The editors and reviewers of this article have no relevant financial relationships to disclose per the Journal policy that requires reviewers to decline review of any manuscript for which they may have a conflict of interest.
References
- 1.Yalon M., Inoue A., Thorne J.E., et al. Infrapopliteal segments on lower extremity CTA: prospective intraindividual comparison of energy-integrating detector CT and photon-counting detector CT. AJR Am J Roentgenol. 2024;222 doi: 10.2214/AJR.23.29778. [DOI] [PubMed] [Google Scholar]
- 2.Gruschwitz P., Hartung V., Kleefeldt F., et al. Photon-counting versus energy-integrating detector CT angiography of the lower extremity in a human cadaveric model with continuous extracorporeal perfusion. Invest Radiol. 2023;58:740–745. doi: 10.1097/RLI.0000000000000982. [DOI] [PubMed] [Google Scholar]
- 3.Willemink M.J., Persson M., Pourmorteza A., Pelc N.J., Fleischmann D. Photon-counting CT: technical principles and clinical prospects. Radiology. 2018;289:293–312. doi: 10.1148/radiol.2018172656. [DOI] [PubMed] [Google Scholar]
- 4.Ren L., Duan X., Ahn R., et al. Photon-counting-detector CT: technology overview and radiation dose reduction. Br J Radiol. 2025;98:1788–1801. doi: 10.1093/bjr/tqaf116. [DOI] [PubMed] [Google Scholar]
- 5.Conte M.S., Bradbury A.W., Kolh P., et al. GVG Writing Group Global Vascular Guidelines on the management of chronic limb-threatening ischemia. J Vasc Surg. 2019;69:3S–125S.e40. doi: 10.1016/j.jvs.2019.02.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Schwartz F.R., Daubert M.A., Molvin L., et al. Coronary artery calcium evaluation using new generation photon-counting computed tomography yields lower radiation dose compared with standard computed tomography. J Thorac Imaging. 2023;38:44–45. doi: 10.1097/RTI.0000000000000685. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Wolf E.V., Halfmann M.C., Schoepf U.J., et al. Intra-individual comparison of coronary calcium scoring between photon counting detector- and energy integrating detector-CT: effects on risk reclassification. Front Cardiovasc Med. 2023;9 doi: 10.3389/fcvm.2022.1053398. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Mergen V., Eberhard M., Manka R., Euler A., Alkadhi H. First in-human quantitative plaque characterization with ultra-high resolution coronary photon-counting CT angiography. Front Cardiovasc Med. 2022;9 doi: 10.3389/fcvm.2022.981012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Emrich T., Hell M. Plaque composition on ultra-high-resolution coronary computed tomography angiography with optical coherence tomography correlation. Eur Heart J. 2023;44:1765. doi: 10.1093/eurheartj/ehac560. [DOI] [PubMed] [Google Scholar]
- 10.Mergen V., Sartoretti T., Baer-Beck M., et al. Ultra-high-resolution coronary CT angiography with photon-counting detector CT: feasibility and image characterization. Invest Radiol. 2022;57:780–788. doi: 10.1097/RLI.0000000000000897. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hartung V., Huflage H., Augustin A.M., et al. Diagnostic limitations in advanced stage peripheral arterial disease in a cadaveric study comparing photon-counting and energy-integrating CT detectors. Sci Rep. 2025;15:6923. doi: 10.1038/s41598-025-91239-x. [DOI] [PMC free article] [PubMed] [Google Scholar]