Head-to-head comparison of ¹⁸F-sodium fluoride coronary PET imaging between a silicon photomultiplier with digital photon counting and conventional scanners

Abstract

Background:

We compared silicone photomultipliers with digital photon counting (SiPM) and photomultiplier tubes (PMT) PET in imaging coronary plaque activity with ¹⁸F-sodium fluoride (¹⁸F-NaF) and evaluated comprehensively SiPM PET reconstruction settings.

Methods:

In 25 cardiovascular disease patients (mean age 67±12 years), we conducted ¹⁸F-NaF PET on a SiPM (Biograph Vision) and conventional PET (Discovery 710) on the same day as part of a prospective clinical trial (NCT03689946). Following administration of 250 MBq of ¹⁸F-NaF, patients underwent a contrast-enhanced CT angiography and a 30-min PET acquisition in list mode on each PET consecutively. Image noise was defined as mean standard deviation of blood pool activity within the left atria. Target-to-background ratio (TBR) and signal-to-noise ratio (SNR) were measured within the whole-vessel tubular 3-dimensional volumes of interest on the cardiac motion and attenuation corrected ¹⁸F-NaF PET images using dedicated software.

Results:

There were significant differences in image noise and background activity between the two PETs (Image noise (%), PMT: 7.6±3.7 vs. SiPM: 4.0±2.3, p<0.001; background activity, PMT: 1.4±0.4 vs. SiPM: 1.0±0.3, p<0.001). Similarly, the SNR and TBR were significantly higher in vessels scanned with the SiPM PET (SNR, PMT: 16.3±11.5 vs. SiPM: 32.7±29.8, p<0.001; TBR, PMT: 0.8±0.4 vs. SiPM: 1.1±0.6, p<0.001). SiPM PET image reconstruction with a 256 matrix, 1.4 mm pixel, and 2 mm Gaussian filter provided best tradeoff in terms of maximal SNR, TBR and clinically practical file size.

Conclusions:

In ¹⁸F-NaF coronary PET imaging, the SiPM PET showed superior image contrast and less image noise compared to PMT PET.

Graphical Abstract

Impact of hardware and reconstruction setup on coronary ¹⁸F-NaF uptake. Coronary ¹⁸NaF PET imaging of a 79-year-old male with coronary ¹⁸F-NaF uptake in the left anterior descending artery. The image noise and background activity were 1.35 %, 0.85 in SiPM PET scanner (SiPM 1), and 5.40 %, 1.33 in PMT PET scanner. The standardized uptake values (SUVmax), signal-to-noise ratio (SNR), target-to-background ratio (TBR) and coronary microcalcification activity (CMA) were 0.91, 63.24, 1.07 and 4.47, respectively in SiPM PET scanner (SiPM 1), and 1.07, 19.81, 0.80 and 4.39, respectively in PMT PET scanner. The SiPM PET scanner images showed less image noise and better image contrast compared to PMT PET scanner images with closed reconstruction parameters. Among three SiPM PET scanner reconstruction settings, the one with a 440 × 440 matrix, slice thickness 1 mm, pixel spacing 0.8mm, and a 2mm Gaussian filter yielded the highest values of SUVmax, SNR, TBR and CMA, while also displaying the worst values for background activity and image noise. ¹⁸F-NaF = ¹⁸F-sodium fluoride; CCTA = coronary computed tomography angiography; SiPM = silicone photomultipliers with digital photon counting; PET = positron emission tomography; PMT = photomultiplier tube

CONDENSED ABSTRACT

This study compares the image quality between silicone photomultipliers with digital photon counting (SiPM) and photomultiplier tubes (PMT) PET in ¹⁸F-sodium fluoride coronary PET. We studied 25 patients with cardiovascular disease who underwent CT angiography and PET/CT acquisitions with two scanners. Image noise, target-to-background ratio and signal-to-noise were measured. There were significant improvements in the aforementioned metrics on SiPM PET imaging (p<0.001). The SiPM PET showed superior image quality compared to the conventional one and a potential usefulness for clinical practice.

INTRODUCTION

¹⁸F-sodium fluoride (¹⁸F-NaF) reflects active calcification in atherosclerotic plaque because it is incorporated directly into exposed hydroxyapatite crystal (1, 2). ¹⁸F-NaF uptake on positron emission tomography (PET) is a marker of developing microcalcification in vascular plaque (1, 3), providing a non-invasive assessment of the biological activity of atherosclerotic plaques in patients with a wide range of cardiovascular diseases, and enabling the prediction of subsequent disease progression and future myocardial infarction (4–9).

Recently, a new generation of PET detectors with silicone photomultipliers (SiPM) has entered clinical use (10–12). According to phantom and patient studies, SiPM PET provides an opportunity for improved image quality and superior lesion detection for oncology patients compared to conventional PET scanners using photomultiplier tubes (PMT) (13). While both types of scanners are currently used in coronary PET clinical trials, the image quality provided by these two types of scanners has not been compared in this technically demanding application.

Therefore, in this study, we aimed to directly compare coronary microcalcification imaging with ¹⁸F-NaF between SiPM and PMT PET, and to optimize the imaging, reconstruction and post-processing of SiPM PET coronary imaging.

MATERIALS AND METHODS

Patient Population

We used data from the ongoing prospective, interventional EVOLVE study (NCT03689946 Effect of Evolocumab on Coronary Artery Plaque Volume and Composition by Coronary CTA and Microcalcification by F18-NaF PET). In Evolve, the effect of evolocumab was assessed using coronary computed tomography (CT) angiography and ¹⁸F-NaF PET before and after 18 months of therapy in established cardiovascular disease patients with noncalcified coronary plaque volume that exceeded 440 mm³ and thoracic aorta atherosclerosis (14). The patients with active atrial fibrillation, pregnancy, and a history of coronary artery bypass graft, etc., were excluded. Patients who underwent two PET scans, using digital and conventional PET, on the same day after ¹⁸F-NaF administration were included in the current analysis. Our study was approved by the Institutional Review Board of the Cedars-Sinai Medical Center and complied with the principles of the Declaration of Helsinki. All patients provided written informed consent prior to their inclusion in this study.

PET and CT acquisition and reconstruction

Figure 1 shows the flow chart of this study. All patients underwent two PET scans on the same day using a SiPM PET scanner (Biograph Vision 600, Siemens Healthineers, Knoxville, Texas, USA) and a PMT PET scanner (Discovery 710, GE Healthcare, Chicago, Illinois, USA). The order of the two tests was randomly selected, and the patients underwent contrast-enhanced CT angiography approximately 120 min after 250 MBq of ¹⁸F-NaF was administered intravenously (SOMATOM Definition Force 192 slice Dual Source CT scanner, Siemens Healthineers, Knoxville, Texas, USA). CT angiography was conducted with the following settings: 250 ms rotation time, 120 kV, and 461–500 mAs, reconstructed with a 0.5 slice thickness and a Bv40 kernel. After a non-contrast CT attenuation correction scan (120 kV, 20 mAs, 1-mm slice thickness on the SiPM PET/CT scanner; 100 kV, 40 mAs, 5-mm slice thickness on the PMT PET/CT scanner), 30-min PET acquisition with electrocardiography (ECG) gating in list mode was performed on each scanner consecutively, 180 min after ¹⁸F-NaF was injected. On the SiPM PET, the ECG-gated list-mode dataset was reconstructed using an Ultra HD algorithm with time-of-flight (TOF). Using 10 cardiac gates, we reconstructed the data on a 256 × 256 matrix (slice thickness 3.27 mm, pixel spacing 1.4 mm) using 4 iterations, 5 subsets without a Gaussian filter and a 440 × 440 matrix (slice thickness 0.8 mm, pixel spacing, 0.8 mm) using 4 iterations, 5 subsets, and a 2 mm Gaussian filter. On the PMT PET, the ECG-gated list-mode dataset was reconstructed using the VUE Point FX-Sharp VPFX-S (TOF-OSEM with resolution recovery) algorithm (15). Using 10 cardiac gates, we reconstructed the data on a 256 × 256 matrix (slice thickness 3.27 mm, pixel spacing 1.6 mm) using 4 iterations, 24 subsets, and a 5 mm Gaussian filter. To assess the image quality, we compared SiPM PET data reconstructed with a 256 × 256 matrix (slice thickness 3.27 mm, pixel spacing 1.4 mm) and a 5 mm Gaussian filter (SiPM 1) to SiPM PET data reconstructed with a 256 × 256 matrix (slice thickness 1.5 mm, pixel spacing 1.4 mm) and a 2 mm Gaussian filter (SiPM 2) and SiPM PET data reconstructed with a 440 × 440 matrix (slice thickness 1.0 mm, pixel spacing 0.8 mm) with a 2 mm Gaussian filter (SiPM 3).

Fig 1. Study outline.

¹⁸F-NaF = ¹⁸F-sodium fluoride; CT = computed tomography; PET = positron emission tomography; SiPM = silicone photomultipliers with digital photon counting; PMT = photomultiplier tubes

Image analysis

Background blood pool activity was defined as the mean activity within the left atrium that was measured using a 3-dimensional spherical volume of interest (Figure 2A). The same volume of interest defined by each patient’s CT angiography was used for quantification of uptake for both the SiPM and the PMT PET images. Image noise was defined as the mean standard deviation of the blood pool activity. The background blood pool activity and image noise were corrected using previously described blood pool correction (Equation 1) (16).

Fig 2. Coronary ¹⁸F-NaF PET image analysis.

3-dimensional spherical volume of interest (green circle) add arrow in the left atria and tubular 3-dimensional volumes of interest on computed tomography angiography image. (A) Volume of interest in the left atrium; (B, C, and D) Volume of interest in left anterior descending coronary artery. ¹⁸F-NaF = ¹⁸F-sodium fluoride; PET = positron emission tomography

$${\text{SUV}}_{\text{background corrected}}={\text{SUV}}_{\text{background}}*{\text{e}}^{-0.004*(60-\text{t})}$$ Equation 1

We automatically extracted whole-vessel tubular 3-dimensional volumes of interest from the CT angiography images using FusionQuant version 1.23.1205 and AutoPlaque version 3.0, software tools (Cedars-Sinai Medical Center, Los Angeles, California, USA) (Figure 2B–D). The same whole-vessel tubular 3-dimensional volumes of interest that were calculated from each patient’s CT angiography results were used for the SiPM and PMT PET images. The whole-vessel maximum standardized uptake value (SUVmax) within the volume of interest was measured. The signal-to-noise ratio (SNR) (SUVmax in the coronary lesion divided by image noise), target-to-background ratio (TBR) (SUVmax in the coronary lesion divided by background activity), and coronary microcalcification activity (CMA) were also evaluated (17, 18). CMA quantifies ¹⁸F-NaF activity across the entire coronary vasculature and is an independent predictor of myocardial infarction (5, 17). All assessments were performed on cardiac motion and attenuation-corrected images using a dedicated software (FusionQuant). The cardiac motion correction technique enables the alignment of all gates to the end-diastolic position and, as a result, allows for the inclusion of all acquired PET counts (19). All PET images were scanned using both SiPM PET and PMT PET. In a previous study, we demonstrated high reproducibility of microcalcification activity measurements (20). In the current study, measurements were repeated in 10 randomly selected patients by two independent readers (H.H. and K.K.) to assess inter-observer reproducibility. Additionally, H.H. repeated measurements in the remaining 15 patients to evaluate intra-observer reproducibility.

Statistical Analysis

Data of continuous variables are expressed as average ± standard deviation. Various image parameters between the two PET scanners were compared using the Wilcoxon matched-pair signed-rank test. Assessment of inter-observer and intra-observer reproducibility was performed using Pearson’s correlation and Bland-Altman analysis among the PET scanner settings. Statistical significance was set at P < 0.05, and all statistical analyses were performed using STATA for Windows (version 17.0; StataCorp LLC, College Station, Texas, USA).

RESULTS

Patient characteristics are presented in Table 1. In the 25 patients, the mean age was 67±12 years and 21 patients (84%) were men. Dyslipidemia was the most prevalent comorbidity affecting 84% (n=21) of the study population. We assessed 75 coronary vessels in 25 patients.

TABLE. 1.

Patient characteristics

	N = 25 (%)
Age (years)	67 ± 12
Male	21 (84)
Obesity (BMI ≥ 25 kg/m2)	15 (60)
Diabetes mellitus	3 (12)
Hypertension	9 (36)
Dyslipidemia	21 (84)
Smoking	7 (28)
Total plaque volume (mm3)	1310.3 ± 498.0
Total non-calcified plaque volume (mm3)	1217.3 ± 491.8
Total calcified plaque volume (mm3)	93.0 ± 121.8

BMI = body mass index

We observed significant differences in image noise and background activity between scanners. Table 2 shows image reconstruction parameters and the image quality values between PMT PET with clinically used reconstruction setting and SiPM PET with three different reconstruction settings. There were significant differences in image noise and background activity between PMT PET and SiPM PET with the closest settings to those of the PMT (SiPM 1) (p<0.001). SNR and TBR were significantly higher in vessels scanned on SiPM PET (p<0.001). CMA per patient was significantly higher in vessels scanned on SiPM PET (p=0.005). On the other hand, there was no significant difference in CMA per vessel between two scanners (p=0.250).

TABLE. 2.

Image quality values between PMT PET scanner and SiPM PET scanner

	PMT	SiPM 1	SiPM 2	SiPM 3	P value
					PMT vs. SiPM 1	PMT vs. SiPM 2	PMT vs. SiPM 3
Matrix size	256	256	256	440
slice thickness (mm)	3.27	3.27	1.5	1
pixel spacing (mm)	1.6/1.6	1.4/1.4	1.4/1.4	0.8/0.8
Gaussian filter (mm)	5	5	2	2
All patients (N=25)
Background activity	1.36 ± 0.39	0.95 ± 0.25	0.94 ± 0.24	1.08 ± 0.36	<0.001	<0.001	<0.001
Image noise, %	7.61 ± 3.74	3.97 ± 2.32	6.34 ± 3.69	7.91 ± 4.61	<0.001	<0.001	0.409
CMA (per patient)	7.81 ± 2.09	8.23 ± 2.80	8.59 ± 3.09	9.08 ± 2.88	0.005	<0.001	<0.001
All vessels (N=75)
SUVmax	1.06 ± 0.70	1.00 ± 0.66	1.30 ± 0.87	1.42 ± 1.39	0.004	<0.001	<0.001
SNR	16.32 ± 11.54	32.73 ± 29.82	29.19 ± 33.56	24.93 ± 31.34	<0.001	<0.001	0.003
TBR	0.78 ± 0.42	1.06 ± 0.60	1.39 ± 0.79	1.27 ± 1.00	<0.001	<0.001	<0.001
CMA (per vessel)	2.60 ± 1.05	2.74 ± 1.36	2.86 ± 1.44	3.03 ± 1.38	0.250	0.030	<0.001

CMA coronary microcalcification activity; SiPM = silicone photomultipliers with digital photon counting; PET = positron emission tomography; PMT = photomultiplier tube; SUV = standardized uptake values; SNR = signal-to-noise ratio; TBR = target-to-background ratio

Among three different SiPM PET reconstruction settings, the approach with SiPM 2 setting demonstrated significantly higher SUVmax, TBR, and per-patient CMA compared to the reconstruction with SiPM 1 setting (all p<0.05) (Figure 3). The setting with SiPM 3 tended to yield higher SUVmax and CMA compared to the setting with SiPM 2 at the expense of higher background activity and image noise (all p<0.001). The reconstruction with SiPM 1 setting showed significantly lower image noise and higher SNR compared to the reconstruction with SiPM 2 setting (all p<0.05) (Figure 3). SiPM 2 setting provided best tradeoff in terms of maximal SNR and TBR. The reproducibility of image quality values between PMT PET scanner and SiPM PET scanner are presented in Supplementary Table 1 and 2.

Fig 3. Quantitative ¹⁸F-NaF uptake and image quality measures on PMT and SiPM PET.

SiPM = silicone photomultipliers with digital photon counting; PET = positron emission tomography; PMT = photomultiplier tubes; CMA = major adverse cardiac events; SUV = standardized uptake values; SNR = signal-to-noise ratio; TBR = target-to-background ratio

DISCUSSION

To our knowledge, this is the first study to compare image quality and quantitative measures of ¹⁸F-NaF coronary uptake on images acquired between a PMT and SiPM PET. In our analysis, the SiPM PET showed significantly lower background activity, and higher SNR and TBR compared to the PMT PET.

Cardiovascular ¹⁸F-NaF PET imaging is a challenging application for PET. Due to the small targets and complex multifactorial motion patterns this advanced imaging technique provides relatively low signal to noise and target to background ratios especially when compared to oncological studies. By means of protocol developments, including prolonged, typically 30-minute acquisitions, ECG gating and postprocessing such as motion correction, ¹⁸F-NaF PET has greatly improved (4, 20–22). The introduction of SiPM PET provides hope that further gains in image quality, diagnostic sensitivity and reproducibility shall be attainable.

In the current study, we explored the added value of SiPM PET and aimed to evaluate the reconstruction setup for digital PET. To ascertain that our comparisons are valid we performed SiPM and PMT emission scanning following one another during a single day. As a result, the acquisitions were performed after a single tracer injection and hence our analysis is not prone to variability in the amount of activity received by the patient. On the other hand, we had to offset the inevitable decrease in background activity, which has been extensively studied in the past (23–25). To address this issue, we used a previously validated conversion factor and normalized the background activities (16). Following this harmonization, we were able to elucidate the potential of SiPM and test how subtle changes in the reconstruction setting affect image quality and the measured ¹⁸F-NaF activity.

Our study showed that the SiPM PET had significantly lower background activity, higher SNR and higher TBR compared to PMT PET. The narrow crystal and the full coverage optimized light collection allows for a high system spatial resolution and improved timing resolution that improved TOF reconstruction algorithm to reduce noise or enhance contrast (26, 27).

There are some practical considerations related to high matrix size with cardiac PET. On SiPM PET, a full study reconstructed with 10 ECG gates results in approximately 1.90 Gigabyte of data which can make it unsuitable for clinical use. This issue will be compounded if respiratory or patient motion needs also be corrected in addition to cardiac motion. As a reasonable compromise, SiPM PET images with a 256 × 256 matrix, slice thickness 1.5 mm, pixel spacing 1.4 mm, 2 mm Gaussian filter, we found significantly reduced background activity and image noise, and significant improvement in SNR and enhancement including SUV max, TBR, CMA per patient and CMA per vessel (Figure 3) as compared to standard protocols on the PMT PET. Therefore, we recommend these reconstruction parameters on the SiPM PET image for routine high-resolution coronary PET imaging.

Study Limitations

This study has several limitations. The number of patients was relatively small, which limited the statistical reliability. While we were not able to provide alternative reconstructions for the PMT PET, as the approach employed in our analysis has been established within a dedicated study and widely adopted (9, 28, 29). However, our results have clearly demonstrated that the SiPM PET showed significantly better image contrast and less image noise than the PMT PET. There are discrepancies in the axial field of view between the SiPM and the PMT PET scanner, which might affect the NEMA sensitivity and contribute to the overall image quality. In our study, we were unable to dissect the impact of individual factors. In the future, studies should aim to critically assess which differences in scanner design and setup have the largest influence on the image quality.

CONCLUSION

In ¹⁸F-NaF coronary PET imaging, the SiPM PET showed significantly better image contrast and a trend toward less image noise compared to PMT PET. Therefore, it may provide better diagnostic performance in the assessment of the coronary microcalcification activity.

New Knowledge Gain and Clinical Implications.

What is new?

• To compare the image quality between silicone photomultipliers with digital photon counting (SiPM) and photomultiplier tubes (PMT) PET in ¹⁸F-sodium fluoride coronary PET.

• SiPM PET provides significantly better image quality compared to PMT PET.

What are the clinical implications?

• The SiPM PET provides superior image quality to the PMT PET and a potential usefulness in medical practice by adjusting settings.

• A reconstruction with a 256 × 256 matrix, slice thickness of 1.5 mm, pixel spacing 1.4 mm, and a 2 mm Gaussian filter provided best tradeoff in terms of maximal SNR, TBR without high disk memory expense.

ABBREVIATIONS

¹⁸F-NaF

¹⁸F-sodium fluoride

SiPM

silicone photomultipliers enabling digital photon counting

PMT

photomultiplier tubes

CT

computed tomography

SUV

standardized uptake values

SNR

signal-to-noise ratio

TBR

target-to-background ratio

CMA

coronary microcalcification activity

TOF

time-of-flight