Photon-Counting Computed Tomography (PC-CT) of the spine: impact on diagnostic confidence and radiation dose

Objectives Computed tomography (CT) is employed to evaluate surgical outcome after spinal interventions. Here, we investigate the potential of multispectral photon-counting computed tomography (PC-CT) on image quality, diagnostic confidence, and radiation dose compared to an energy-integrating CT (EID-CT). Methods In this prospective study, 32 patients underwent PC-CT of the spine. Data was reconstructed in two ways: (1) standard bone kernel with 65-keV (PC-CTstd) and (2) 130-keV monoenergetic images (PC-CT130 keV). Prior EID-CT was available for 17 patients; for the remaining 15, an age–, sex–, and body mass index–matched EID-CT cohort was identified. Image quality (5-point Likert scales on overall, sharpness, artifacts, noise, diagnostic confidence) of PC-CTstd and EID-CT was assessed by four radiologists independently. If metallic implants were present (n = 10), PC-CTstd and PC-CT130 keV images were again assessed by 5-point Likert scales by the same radiologists. Hounsfield units (HU) were measured within metallic artifact and compared between PC-CTstd and PC-CT130 keV. Finally, the radiation dose (CTDIvol) was evaluated. Results Sharpness was rated significantly higher (p = 0.009) and noise significantly lower (p < 0.001) in PC-CTstd vs. EID-CT. In the subset of patients with metallic implants, reading scores for PC-CT130 keV revealed superior ratings vs. PC-CTstd for image quality, artifacts, noise, and diagnostic confidence (all p < 0.001) accompanied by a significant increase of HU values within the artifact (p < 0.001). Radiation dose was significantly lower for PC-CT vs. EID-CT (mean CTDIvol: 8.83 vs. 15.7 mGy; p < 0.001). Conclusions PC-CT of the spine with high-kiloelectronvolt reconstructions provides sharper images, higher diagnostic confidence, and lower radiation dose in patients with metallic implants. Key Points • Compared to energy-integrating CT, photon-counting CT of the spine had significantly higher sharpness and lower image noise while radiation dose was reduced by 45%. • In patients with metallic implants, virtual monochromatic photon-counting images at 130 keV were superior to standard reconstruction at 65 keV in terms of image quality, artifacts, noise, and diagnostic confidence.


Introduction
Surgical treatment of the spine is recommended as first-line therapy for various conditions including unstable fractures and neoplastic diseases as well as severe degeneration and is expected to further increase due to demographic changes of the population [1]. Frequently, the use of metallic implants is necessary for successful treatment [2]. Computed tomography (CT) is the imaging modality of choice to assess postsurgical outcome and potential complications of both the implant and the adjacent anatomy [2]. However, artifacts caused by the implants constitute a major challenge for accurate assessment. These artifacts have a characteristic, streak-like appearance and are caused by various physical processes, such as photon scattering, photon starvation, and beam hardening [3], with the degree of these different physical processes depending on the type and density of the material. As a result, the shift in the x-ray energy spectrum and the lack of photons arriving at the detector can lead to a substantially impaired image quality including reduced soft tissue contrast, increased image noise, or an entire signal loss posing the risk of missing relevant findings [4].
To overcome these limitations, various approaches to reduce metallic artifacts have been developed and implemented in clinical routine. The most widely used techniques comprise calculation of virtual monochromatic images at high kiloelectronvolt levels from dual-energy CT data [5,6] or the use of dedicated metal artifact reduction (MAR) algorithms based on sinogram inpainting, frequency-split techniques, iterative reconstruction approaches, and artificial intelligence to effectively reduce artifacts [7][8][9][10][11][12].
With the recent introduction of photon-counting CT (PC-CT) into clinical routine, a novel technology for data acquisition and reconstruction has become available with the potential to fundamentally change current workflows [13]. PC-CT allows for acquiring multispectral image information in every scan, which can be binned at specific kiloelectronvolt ranges and used for further post-processing and quantitative image analysis [13]. Although the full potential of this technique is still in its infancy, exploiting the multispectral data may be helpful to reduce artifacts from metallic implants and improve diagnostic confidence at a low radiation dose [14][15][16].
As mainly preclinical phantom-study data are available, we aimed to investigate the clinical impact of PC-CT compared to conventional energy-integrating CT (EID-CT) for spinal imaging with respect to image quality, diagnostic confidence, and radiation dose. We hypothesize that multispectral PC-CT data with reconstruction of high-kiloelectronvolt images allow for improved diagnostic confidence at reduced radiation dose compared to the currently established clinical standard.

Ethics approval
In this prospective study approved by the Institutional Review Board (Ethics Committee, University of Freiburg), written informed consent was obtained from all patients. All procedures were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Patient cohort
We prospectively enrolled patients with clinically indicated non-enhanced PC-CT of the spine between November 2021 and March 2022. An EID-CT of the same scan region served as reference standard, which was either available from prior CT imaging of the same patient or as a matched case with regard to age, sex and body mass index (BMI). BMI was chosen as a matching parameter as it is related to the bone density [17] which itself might influence the subjective image impression.

CT imaging acquisition and reconstruction
PC-CT All PC-CT examinations were performed on a NAE-OTOM Alpha (Siemens Healthineers) in supine position and without additional contrast agent. Acquisition parameters were as follows: multispectral QuantumPlus mode, CARE Dose4D & CARE kiloelectronvolts set to manual kilovolts, tube voltage 120 kV, CARE kiloelectronvolt IQ level 85 (corresponding to effective mAs 60), rotation time 0.5 s, pitch 0.8, focal spot 0.8 × 1.2 mm, and standard collimation of 144 × 0.4 mm).
From the acquired multispectral data, a standard series (PC-CT std ) was reconstructed at 65 keV using a dedicated bone kernel (Br64) (slice thickness of 2 mm, an increment of 2 mm) and quantum iterative reconstruction (QIR) with 1 3 strength 2. These protocol settings were determined in an evaluation period preceding the current study to model the clinical standard EID-CT protocol and maintain image impression at reduced radiation dose.
In addition, a second series of virtual monoenergetic images at 130 keV was reconstructed inline exploiting the multispectral PC-CT data (PC-CT 130 keV ). All other reconstruction parameters were kept similar.
EID EID-CT examinations were performed on a thirdgeneration CT SOMATOM Definition AS (Siemens Healthineers). Images were acquired using the standard clinical protocol with single source with CARE Dose4D & CARE kiloelectronvolts set to manual kilovolts, tube voltage 120 kV, rotation time 1.0 s, pitch 0.8, focal spot 0.9 × 1.2 mm, and standard collimation of 64 × 0.6 mm. Image data was reconstructed with a Br62 standard kernel with 2 mm slice thickness, 2 mm increment, and iterative reconstruction strength 3.
All reading sessions were performed on a dedicated workstation using the publicly available post-processing platform NORA (www. nora-imagi ng. com). Image data were provided to the readers in 2-mm reconstructions in axial, coronal, and sagittal orientations as well as in axial thin-slice image data (0.6 mm for PC-CT and 0.75 mm for EID-CT).
In a subsequent reading session, PC-CT std and PC-CT 130 keV images were evaluated by the same four radiologists in the subset of patients with metallic spinal implants in a similar fashion and blinded to the type of reconstruction.

Quantitative image analysis
Region-of-interest (ROI) measurements were performed to objectify the effectiveness of PC-CT 130 keV for metal artifact reduction compared to the PC-CT std reconstruction. For this purpose, a 10-mm 2 circular ROI was placed in the area with the subjectively most severe artifact appearance caused by photon starvation in PC-CT std . ROI size and location were kept similar for the corresponding measurements in the PC-CT std reconstruction. In addition, a second ROI (10 mm 2 ) was placed in a distant muscle as far away as possible from the implant and in an area subjectly not affected by the artifact serving as reference.

3
An image example is given in Fig. 2(A/B). The mean Hounsfield unit (HU) values from these ROIs were read for PC-CT std and PC-CT 130 keV .

Radiation dose
All dose calculations were performed using a commercially available dose management and reporting platform (DoseM, INFINITT Europe GmbH). Volume CT dose index (CTDI vol [mGy]) and effective doses (mSv) were extracted for the entire cohort and cervical and lumbar scans separately and compared between the PC-CT and EID-CT examinations.

Statistical analysis
All statistical analyses were performed using R statistics version 3.6.3 (R Core Team, https:// www.R-proje ct. org). We used the Shapiro-Wilk statistic to test for normal distribution. Continuous variables are reported as mean and standard deviation or median and ranges/interquartile ranges (IQR) as appropriate. Paired-sample t tests were conducted to compare age, radiation dose, and ROI HU values. Sex distribution between groups was tested with chi-square test. Differences in median reading scores were compared with the Wilcoxon rank sum test based on the average reading scores of the four readers per reading item. Values with an α level of 0.05 were considered statistically significant. p values are descriptive due to the exploratory study design.

Patient characteristics
Spinal PC-CT was successfully performed in all 32 patients (mean age 58.8 ± 15.6 years; 18 [56%] male) included in this study. Among those, prior EID-CT of the spine was available for 17 patients; for the remaining 15 patients, a matched EID-CT cohort with similar baseline demographics could be identified (age, height, weight, and BMI [all p > 0.05]). In n = 10 patients, metallic implants were present. Patients with metallic implants were more likely men (67%), heavier, and older. Detailed patient characteristics are provided in Table 1, which were manually extracted from the electronic medical records.
In the subset of patients with metallic implants, reading scores for PC-CT 130 keV revealed significantly superior  (Fig. 4). A summary of all reading scores is provided in Table 3.

Quantitative assessment of image quality
Quantitative measurements revealed a significant artifact reduction for PC-CT 130 keV compared to PC-CT std . Mean HU values of the ROI within the subjectively most severe artifact were significantly lower in the PC-CT std vs. the PC-CT 130 keV reconstruction (− 462 ± 191 HU vs. 26 ± 108 HU, respectively; p < 0.001). No significant differences were found in the measurements of the reference ROI placed in the distant muscle subjectively not affected by the artifact (48 ± 19 HU vs. 46 ± 15 HU; p = 0.575; Fig. 2).

Radiation dose
Radiation dose analyses revealed a significantly lower CTDI vol for PC-CT compared to EID-CT with a mean CTDI vol of 8.83 vs. 15.7 mGy; p < 0.001 for the entire cohort. Similar results were found when looking at the

Discussion
In this study, we investigated the clinical impact of PC-CT compared to conventional EID-CT for spinal imaging with respect to image quality, diagnostic confidence, and radiation dose. We found that PC-CT enables a significantly improved edge sharpness and less image noise compared to the current clinical standard while radiation dose can be substantially reduced (CTDI vol by 45%; effective dose by 65%). In addition, inline reconstruction of high-kiloelectronvolt monoenergetic reconstructions at virtual 130 keV provides significantly higher diagnostic confidence in patients with metallic implants while metal artifacts were effectively reduced. These results are of clinical importance as CT imaging of the spine is routinely performed in the workup of patients with chronic back pain, trauma, and oncological diseases for both treatment planning and follow-up [18,19]. As metallic implants are frequently necessary for successful treatment, they are encountered on a considerable number of scans [2]. Despite substantial advances in recent years, artifacts caused by such implants still pose diagnostic challenges with the risk for delayed diagnosis of implant failure and associated complications. The investigated approach using a high-kiloelectronvolt monoenergetic reconstruction has emerged as a reliable and efficient way for artifact reduction [20][21][22]. Laukamp et al. found that virtual monoenergetic reconstructions at 200 keV allowed for a significant artifact reduction in patients with hip replacement [23]. In addition, Hokamp et al. reported that high-kiloelectronvolt monoenergetic images can effectively reduce artifacts of dental hardware [24]. In line with these results, we found that high-kiloelectronvolt reconstructions at 130 keV significantly reduce artifacts and  Images depict an axial slice at the height of the 5th cervical vertebra. EID-CT shows severe artifacts caused by metallic implants resulting in a substantially impaired image quality and reduced diagnostic confidence. Screw loosening was suspected in the left pedicle (arrow). After treatment of a delayed wound healing, follow-up CT imaging was performed on a PC-CT after interim external immobilization of the neck. PC-CT std shows significantly reduced artifacts and better depiction of the bone resorption and also reveals loosening of the right screw (arrow). PT-CT 130 keV reconstruction shows almost no artifacts and clearly depicts the bone resorption around both screws (arrows). EID-CT = energyintegrating CT; PC-CT std = photon-counting CT; PT-CT 130 keV PC-CT with monoenergetic reconstructions at 130 keV increase diagnostic confidence in patients with metallic spinal implants. Up till now, dedicated dual-energy CT scan modes were necessary for this approach which had to be chosen in advance of the scan. With the recent introduction of PC-CT into clinical routine, a new method has become available, which allows for acquiring the necessary information in every scan by using the multispectral properties of the new detector technology [13]. From data acquired with the QuantumPlus mode, this facilitates reconstruction of virtual monoenergetic images at arbitrary kiloelectronvolt levels between 40 and 190 keV at a significantly reduced radiation dose. Besides virtual high-kiloelectronvolt monoenergetic reconstructions, dedicated algorithms have been proposed to effectively reduce artifacts caused by metallic implants. These algorithms have shown promising results for various types of implants such as dental hardware and hip prosthesis [25][26][27], especially when combined with high-kiloelectronvolt monoenergetic reconstruction [3,25,28,29]. When the data for this study was collected, dedicated metal artifact reduction software was not clinically available for the investigated PC-CT system. This has changed in the meantime and will be evaluated in future work.
The new PC-CT detector technology not only enables acquisition of multispectral data but also has a substantially higher resolution compared to current state-of-the-art EID-CT systems [30,31]. This is of particular benefit for the depiction of subtle structures such as trabecular bone [32]. Preliminary results have shown superior spatial resolution and delineation of the paranasal sinuses and temporal bone at a significantly reduced dose compared to a conventional EID-CT scanner [33,34]. Our results also revealed a significantly higher edge sharpness of osseous structures/details in comparison to the current clinical standard while significantly reducing radiation dose. As the PC-CT protocol was set up to mimic the current scanner in terms of subjective image impression, an even higher dose reduction seems possible when fully exploiting the new detector technology and higher dose efficiency [35].
The following limitations of this study need to be considered. First, no detailed information (such as composition or alloy) of the metallic implants was available. Thus, the impact of different implant types on image quality could not be further investigated. Second, no dedicated metal artifact reduction algorithms were investigated as they were not yet clinically available for PC-CT. Further, the sample size, especially of patients with metallic implants, was relatively small. Future prospective investigations are necessary to confirm our findings. Finally, the EID-CT scans were acquired on a scanner without a dual-source detector; thus, virtual monochromatic data was not available. Despite thorough blinding of the readers, identification of PC-CT scans vs. EID-CT scans based on image sharpness and resolution cannot be finally excluded. Lower radiation dose exposure and improved image quality can be assumed for more recent, third-generation scanners even though our comparison may more realistically represent the clinical routine given their broader availability.
In conclusion, PC-CT of the spine with high-kiloelectronvolt monoenergetic reconstructions provides significantly sharper images and higher diagnostic confidence in patients with metallic implants and enables effective artifact reduction at a low radiation dose. This is of clinical importance, as follow-up imaging is frequently performed, and early diagnosis of post-surgical complications (implant failure, screw loosening) is important to initiate appropriate interventions.