Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Jun 7.
Published in final edited form as: Proc SPIE Int Soc Opt Eng. 2022 Apr 4;12031:120311R. doi: 10.1117/12.2612999

Quantification of Coronary Calcification using High-Resolution Photon-Counting-Detector CT and an Image Domain Denoising Algorithm

Patrick VanMeter 1, Jeffrey Marsh Jr 1, Kishore Rajendran 1, Shuai Leng 1, Cynthia McCollough 1
PMCID: PMC9172081  NIHMSID: NIHMS1810809  PMID: 35677470

Abstract

Coronary artery calcification (CAC) is an important indicator of coronary disease. Accurate volume quantification of coronary calcification, especially calcifications smaller than a few mm, using computed tomography (CT) is challenging due to calcium blooming, which is a consequence of limited spatial resolution. In this study, ex-vivo coronary specimens were scanned on a clinical photon-counting detector (PCD) CT scanner and the estimated coronary calcification volume were compared with a conventional energy-integrating detector (EID) CT. Scans were performed using the same tube potential and radiation dose (120 kV, 9.3 mGy CTDIvol). EID-CT images were reconstructed using our routine clinical protocol for CAC quantification. PCD-CT images were reconstructed using a sharper reconstruction kernel than that was supported by the EID-CT scanner, resulting in improved resolution but higher image noise levels. An image-based denoising algorithm was applied to the PCD-CT images to achieve similar noise levels as the EID-CT images. Calcifications were segmented to estimate the volume. Micro-CT images of the same calcifications were acquired and served as the reference standard. PCD-CT images showed reduced calcium blooming artifacts compared to EID-CT. Calcification volume estimates were found to overestimate the micro-CT volumes by 9 ± 12% for PCD-CT data, and 24 ± 18% for the EID-CT data. Volume quantification accuracy of the current PCD-CT system was also found to be superior to a previous-generation investigational PCD-CT scanner with larger detector pixels.

Keywords: Photon counting detector CT, coronary artery disease, coronary calcifications, calcium quantification, image domain denoising

1. INTRODUCTION

The development of CAC is known to be correlated with the progression of coronary artery disease and atherosclerosis [1–4]. The size of coronary calcifications and degree of luminal stenosis are commonly assessed using coronary computed tomography angiography procedures [5–7]. Calcified deposits within vessel walls present as brighter objects in CT images due to their high attenuation. This characteristic enables volume quantification of the calcifications through image segmentation. However, the accuracy of coronary calcification volume estimation in conventional EID-CT is limited due to the blooming artifacts caused by insufficient spatial resolution [8, 9]. On the contrary, PCD-CT has demonstrated higher spatial resolution at routine dose [10], which could reduce the blooming effect and consequently improve volume estimation accuracy [11].

Prior work by our group demonstrated that coronary calcification blooming artifacts were reduced using a previous-generation research PCD-CT system (SOMATOM CounT, Siemens Healthcare GmbH) compared to its complementary second-generation EID-CT (SOMATOM Definition Flash; Siemens Healthcare GmbH) [11]. All volumetric comparisons were performed on calcifications residing within excised human cadaveric coronary artery specimens submerged in an anthropomorphic water phantom. Segmentation of the selected calcifications in contiguous CT images revealed a lower volume using PCD-CT for 12/13 specimens assessed.

The purpose of the current study is to investigate further improvements in calcium volume quantification using a new generation clinical high-resolution PCD-CT scanner in comparison with a conventional state-of-the-art EID-CT scanner. The same human cadaveric specimens were used in this application for consistency with previous work. This enabled further comparison of the calcification volumes obtained in the current study with those obtained using the previous-generation PCD-CT system (SOMATOM CounT).

2. METHODS

2.1. Specimen preparation and calcification identification

Six coronary arteries and one coronary venous graft were excised from three human cadavers, fixed in neutral-buffered formalin, and embedded in methyl methacrylate (MMA) blocks. Calcifications were identified within these ex vivo specimens by a thoracic radiologist with 11 years of cardiac CT experience [11]. A total of 12 calcifications were identified among the coronary artery specimens, and an additional calcification was included from the venous graft, bringing the total to 13 distinct specimens.

2.2. Data acquisition and image reconstruction

The coronary artery specimens were placed into a 30 cm body-shaped water tank and scanned on a commercial third-generation dual-source EID-CT scanner (SOMATOM Force, Siemens Healthcare) and a clinical full-body dual-source PCD-CT scanner (NAEOTOM Alpha, Siemens Healthcare) built on the same platform as the EID-CT [12]. The PCD-CT scan was acquired using a high-resolution (HR) mode which features 150 x 176 μm2 pixels (at iso-center) and one energy threshold. Radiation dose was matched between the two scans (CTDIvol = 9.3 mGy), and the same tube potential was used (120 kV).

Images were reconstructed using each platform’s native iterative reconstruction technique. EID-CT images were reconstructed using a routine quantitative kernel (Qr49; 0% MTF at 12.4 lp/cm), following standard protocol in our practice, while the high resolution PCD-CT data were reconstructed with a sharp quantitative kernel (Qr68; 0% MTF at 15.5 lp/cm) that is only available on the PCD-CT.

To establish a reference volume, each of the coronary specimens were individually imaged on a micro-CT scanner featuring an effective pixel pitch of 20 μm. Micro-CT images were reconstructed using an in-house Feldkamp algorithm. The corresponding micro-CT volume estimations were treated as the reference, with which the measurements from PCD-CT and EID-CT were compared.

2.3. PCD-CT image domain denoising

Image noise was estimated by measuring the standard deviation within circular regions of interest positioned in the water surrounding the calcification specimens. The sharper quantitative kernel used to produce the PCD images was better suited to resolve the high frequency boundaries captured by the higher resolution PCD detector, however, it also increased high frequency noise. Thus, an image domain denoising algorithm was used to limit the influence of high frequency noise on the ensuing PCD-CT calcification volume measurements. A prior knowledge-based noise-reduction algorithm, Slice-PKAID [13], was therefore applied to the PCD-CT data to achieve a noise level similar to that observed in the EID-CT images.

2.4. Segmentation and volume estimation

Segmentation was performed using a half-maximum thresholding (HMT) methodology [11, 14]. This methodology is designed to minimize the impact of blooming on volume estimation for highly attenuating objects with a relatively uniform composition. For each slice, the user manually positioned a small circular region of interest (ROI) onto the most attenuating region of the calcification. Another ROI was placed in the background MMA. The average CT number (HU) was calculated within each ROI, and the threshold for HMT segmentation was selected to be the halfway point between these two averages. The total volume was calculated as the product of the individual voxel size (determined by the reconstruction field of view and image matrix size) and the number of segmented voxels. The HMT segmentation procedure is illustrated in Figure 1.

Figure 1.

Figure 1.

Open in a new tab

Schematic illustrating the segmentation process. (a) The original PCD-CT image of a sample calcification specimen (23.4×23.4 mm field of view). The ROI selections for calcification and background used for determining the HMT are shown. Window level/width: 398/995 HU. (b) Segmentation mask displaying pixels above the HMT in white. (c) Overlay of the segmentation mask onto the CT image. The segmented pixels are highlighted using a color gradient, with a brighter color corresponds to a greater CT number.

2.5. Previous-generation PCD-CT volume measurements

The calcification volume measurements reported herein could be further compared with results reported in earlier work by our group, which examined the same specimens using a previous-generation PCD-CT scanner [11]. The PCD-CT scanner used in that work was an investigational PCD-CT system (SOMATOM CounT, Siemens Healthcare) which was built upon a modified dual-source SOMATOM Flash system [15]. The scanner featured an isocenter detector pixel size of 250 × 250 μm2 and had a limited 275 mm field-of-view.

2.6. Statistical analysis

The percent overestimation of the PCD-CT and EID-CT volume estimates relative to micro-CT was determined using linear least squares regression (VCT = β1 VμCT + β0 + ϵ), with 100% × (β1 – 1) interpreted as the percent overestimation (i.e., 1.10 is a 10% overestimation). The results are reported as 95% confidence intervals based on the Student’s t-distribution with N – 2 degrees of freedom, where N is the number of calcifications. Image noise and contrast-to-noise ratio (CNR) were quantified using the ROI measurements from the calcium and background.

3. RESULTS

3.1. Impact of PCD-CT denoising

The image noise within the Qr49 EID-CT images was measured as 42.6 ± 0.9 HU. Conversely, the image noise within the original Qr68 PCD-CT images was measured as 74.0 ± 1.3 HU. After application of the Slice-PKAID denoising algorithm, the PCD-CT image noise was reduced to 34.8 ± 0.9 HU, which was comparable to that of the EID-CT (Figure 2).

Figure 2.

Figure 2.

Open in a new tab

Comparison of noise texture and magnitude in a 128 x 128 pixel section of the water surroundings between (a) PCD-CT (original), (b) PCD-CT (denoised), and (c) EID-CT. Noise magnitude is similar in (b) and (c) but noise texture is not identical.

3.2. Calcification volume assessment

For all 13 specimens, the PCD-CT images resulted in a smaller volume estimation compared to EID-CT. In 12/13 cases, the PCD-CT estimation is closer to the micro-CT reference volume, while in the remaining case PCD-CT underestimates micro-CT by approximately the same amount as EID-CT overestimates. The mean CNR was 26.56 ± 2.30 for the PCD-CT images and 12.43 ± 1.14 for EID-CT. PCD-CT volume estimates were found to overestimate the reference volume by 9 ± 12% while EID-CT overestimates by 24 ± 18% (95% CI). Figure 3 shows a typical comparison between micro-CT, PCD-CT, and EID-CT for a single calcification specimen. Parts (b - c) display the segmented pixels with a superimposed color gradient, where brighter colors correspond to greater CT numbers. A reduced regimented volume and a more uniform CT number distribution can be seen in the PCD-CT image, indicating reduced blooming.

Figure 3.

Figure 3.

Open in a new tab

Comparison between (a) micro-CT, (b) PCD-CT, and (c) EID-CT image segmentations for a single axial slice of a sample specimen with a 7.5 x 7.5 mm field of view. Measured volumes are 45.0 mm3, 49.2 mm3, and 64.9 mm3 respectively. The PCD-CT image shows significantly decreased blooming, especially in regions with small structures. The PCD-CT image also correctly reveals that the calcium does not form a closed loop, which cannot be determined in the EID-CT image.

3.3. Comparison to previous-generation PCD-CT

The current-generation PCD-CT scanner obtains smaller and more accurate volume measurements for 12/13 specimens compared to the previous-generation PCD-CT, and the remaining specimen differs by only 0.7 mm3. The regression analysis was repeated for the previous-generation PCD-CT results, which was found to overestimate the reference volume by 18 ± 12% (95% CI). This is superior to the performance of the EID-CT scanner, but inferior to the current-generation PCD-CT scanner (9 ± 12%). A graphical comparison of the volume measurement accuracy for all three scanners is shown in Figure 4.

Figure 4.

Figure 4.

Open in a new tab

Calcification volume, as a percentage of the micro-CT volume, measured with the EID-CT and two PCD-CT scanners (the previous-generation “CounT” and the current-generation “Alpha”). The dashed red line designates 100% of the micro-CT reference volume.

4. DISCUSSION

Evaluation of CAC volume using a clinical PCD-CT system demonstrated increased accuracy compared to EID-CT, relative to the micro-CT reference. The PCD technology enables deployment of smaller detector pixels which allows CT scans to attain higher resolution than conventional EID-CT. Although the size of EID-CT detector elements can also be decreased, this would reduce the detector geometric efficiency (fill factor), considering that the thickness of septa cannot be substantially reduced. Conversely, PCDs do not require interpixel septae and can therefore accommodate smaller detector cells without losing geometric efficiency.

Sharper convolution kernels are needed for PCD-CT reconstructions to take advantage of the higher resolution scan data. While the sharp kernels can resolve the high frequency information at structure boundaries, they also tend to inflate high frequency noise. Such image noise can be reasonably reduced by deployment of an image denoising algorithm. Here we showed that the Slice-PKAID image domain denoising algorithm was able to reduce the high frequency noise in the PCD-CT images down to the same levels measured in the EID-CT images.

In addition to being more accurate than the state-of-the-art EID-CT system, the CAC volume estimates obtained from the clinical PCD-CT system were also more accurate than a previous-generation PCD-CT system. Previous-generation PCD technology featured detector pixel sizes of 250 x 250 μm2 (at isocenter), whereas the current generation features pixel sizes of 150 x 176 μm2, yielding a reduction in detector element area of 58%. The improved resolution enabled by the 58% smaller pixel sizes contributed to reducing the volume estimation error from 18 ± 12% down to 9 ± 12%.

5. CONCLUSIONS

This study demonstrates that scanning CACs on a clinical PCD-CT yields reductions in calcium blooming artifacts and improvements in calcification volume quantification accuracy beyond what is achievable with state-of-the-art commercial EID-CT. These results are attributed to the improved spatial resolution of the PCD-CT system which is possible through deployment of smaller sized detector elements.

ACKNOWLEDGMENTS

Research reported in this work was supported by the National Institutes of Health under award number R01 EB028590 and C06 RR018898. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Health. This work was supported in part by the Mayo Clinic X-ray Imaging Research Core.

REFERENCES

  • [1].Rifkin RD, et al. “Coronary calcification in the diagnosis of coronary artery disease.” Am. J. Cardiol. 44:141–147 (1979). [DOI] [PubMed] [Google Scholar]
  • [2].Liu W, et al. “Current understanding of coronary artery calcification.” J. Geriatr. Cardiol. 12(6):668–675 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [3].Sangiorgi G, et al. “Arterial calcification and not lumen stenosis is highly correlated with atherosclerotic plaque burden in humans: a histologic study of 723 coronary artery segments using nondecalcifying methodology.” J. Am. Coll. Cardiol. 31:126–133 (1998). [DOI] [PubMed] [Google Scholar]
  • [4].Rumberger JA et al. “Coronary artery calcium area by electron-beam computed tomography and coronary atherosclerotic plaque area. A histopathologic correlative study.” Circulation. 92:2157–2162 (1995). [DOI] [PubMed] [Google Scholar]
  • [5].Goff DC Jr, et al. “2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines.” J. Am. Coll. Cardiol. 63:2935–2959 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [6].Knuuti J, et al. “2019 ESC Guidelines for the diagnosis and management of chronic coronary syndromes.” Eur. Heart J. 41:407–477 (2020). [DOI] [PubMed] [Google Scholar]
  • [7].Fihn SD, et al. “2012 ACCF/AHA/ ACP/AATS/PCNA/SCAI/STS guideline for the diagnosis and management of patients with stable ischemic heart disease: a report of the American College of Cardiology Foundation/American Heart Association task force on practice guidelines, and the American College of Physicians, American Association for Thoracic Surgery, Preventive Cardiovascular Nurses Association, Society for Cardiovascular Angiography and Interventions, and Society of Thoracic Surgeons.” Circulation. 126:e354–e471 (2012). [DOI] [PubMed] [Google Scholar]
  • [8].Budoff MJ, et al. “Diagnostic performance of 64-multidetector row coronary computed tomographic angiography for evaluation of coronary artery stenosis in individuals without known coronary artery disease: results from the prospective multicenter ACCURACY trial.” J. Am. Coll. Cardiol. 52:1724–1732 (2008). [DOI] [PubMed] [Google Scholar]
  • [9].Abdulla J, et al. “Influence of coronary calcification on the diagnostic accuracy of 64-slice computed tomography coronary angiography: a systematic review and meta-analysis.” Int. J. Cardiovasc. Imaging 28:943–953 (2012). [DOI] [PubMed] [Google Scholar]
  • [10].Leng S, et al. “Photon-counting Detector CT: System Design and Clinical Applications of an Emerging Technology.” RadioGraphics 39(3), 729–743 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [11].Sandstedt M, et al. “Improved coronary calcification quantification using photon-counting-detector CT: an ex vivo study in cadaveric specimens.” Eur. Radiol. 10.1007/s00330-021-07780-6 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [12].Rajendran K et al. “First Clinical Photon-counting Detector CT System: Technical Evaluation.” Radiology. 15:212579. doi: 10.1148/radiol.212579. Epub ahead of print. (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [13].Tao S, et al. “Noise reduction in CT image using prior knowledge aware iterative denoising.” Phys. Med. Biol. 10.1088/1361-6560/abc231 (2020) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [14].Marsh J, et al. “A blooming correction technique for improved vasa vasorum detection using an ultra-high-resolution photon-counting detector CT.” Proc. SPIE 11312, Medical Imaging 2020: Physics of Medical Imaging, 113124R (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [15].Yu Z et al. , “Evaluation of conventional imaging performance in a research whole-body CT system with a photon-counting detector array,” Phys Med Biol. 61(4): 1572–1595 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES