Spectral Imaging With Dual-layer Spectral Detector Computed Tomography for Acute Coronary Syndrome

Purpose: To evaluate the feasibility of spectral imaging with dual-layer spectral detector computed tomography (CT) for the diagnosis of acute coronary syndrome. Methods: We identied 33 consecutive patients who underwent cardiac CT using dual-layer spectral detector CT and were diagnosed with acute ischemic syndrome by an invasive coronary angiography. We reconstructed 120 kVp images and generated virtual monochromatic images (VMIs; 40–200 keV in 10 keV increments), iodine concentration maps, and effective atomic number (Z) maps. We calculated the contrast and contrast-to-noise noise ratio (CNR) between myocardial normal and hypo-perfusion and chose the VMIs with the best CNR for quantitative analysis. We compared the image noise, contrast, and CNR of 120 kVp images and the best VMIs, CT value, iodine concentration, and effective Z between myocardial normal and hypo-perfusion with the paired t-test. Results: As the X-ray energy decreased, venous attenuation, contrast, and CNR gradually increased. The 40 keV image yielded the best CNR. There was no signicant difference in image noise between the 120 kVp and 40 keV images. The contrast and CNR between myocardial normal and hypo-perfusion were signicantly higher in 40 keV images than those in 120 kVp images. The iodine concentration and the effective Z were signicantly higher in normal myocardium than those in hypo-perfused myocardium. Conclusion: Spectral imaging with dual-layer spectral detector CT is a feasible technique to detect the hypo-perfused area of acute ischemic syndrome.


Introduction
Coronary thrombosis is generally accepted as the direct cause of acute coronary syndrome (ACS), and most coronary thrombi occur due to rupture or stulation of the coronary plaque [1,2]. Many studies have suggested that coronary computed tomography (CT) angiography has a high sensitivity and negative predictive value for signi cant coronary stenosis [3]. However, previous reports have also suggested that the percentage of coronary stenosis only modestly correlates with decreased myocardial blood ow [4].
Additionally, serial coronary angiography demonstrated that plaque rupture often occurs at sites with less severe stenosis [5,6]. Therefore, coronary stenosis detected by coronary CT angiography might be inadequate to diagnose the culprit vessel for ACS, as many coronary arteries with intermediate stenosis (50-70%) can undergo unnecessary additional testing and treatments. A previous study suggested that rest CT perfusion by routine cardiac CT scan can increase the positive predictive value of ACS; however, the sensitivity of this method is only 33 % (3/9 patients) [7].
Dual-energy CT (DECT) can create virtual monochromatic images (VMIs) at different monochromatic Xray energies (keV) based on the two different energy datasets. The reported advantages of VMIs include reducing beam-hardening artifacts and providing more accurate quantitative attenuation measurements [8,9]. VMIs at low keV can increase contrast enhancement compared with the conventional tube technique [10]. However, no studies have reported the usefulness of rest cardiac DECT for ACS. Recently, dual-layer DECT (DL-DECT) has been introduced for clinical use [11][12][13]. DL-DECT can overcome the disadvantage of increased image noise at low-energy levels. A previous study reported that VMIs at lower energy levels increased image quality of myocardial late iodine enhancement [14]; however, to the best of our knowledge, no previous study has compared VMIs obtained with DL-DECT with the images obtained with a conventional 120 kVp protocol to evaluate ACS.
The purpose of this study was to evaluate the feasibility of spectral imaging with dual-layer spectral detector CT for the diagnosis of ACS.

Methods
This retrospective study received institutional review board approval; the requirement for written informed consent was waived.

Study population
This was a retrospective study of the clinical records of ACS patients who underwent cardiac CT at our institution between January 2017 and March 2018. In this study period, coronary CT angiography was performed on 3177 patients with suspected or con rmed coronary artery disease for clinical reasons, based on the guidance of the American College of Cardiology [15]. Among these patients, 46 patients were diagnosed with ACS by invasive coronary angiography, of which 13 patients were excluded because of different CT protocols (n=9) or poor image quality (n=4). The remaining 33 patients were enrolled in this study and included 22 men and 11 women with a mean age of 62 years (range: 42-91 years) and a mean body weight of 60.3 kg (range: 32-99 kg).

CT Image Reconstruction
The spectral-based image data were post-processed at a workstation (Spectral Diagnostic Suite; Philips Healthcare) to generate VMIs at 17 different energy levels (40-200 keV) with a spectral level of 3 (manufacturer's recommendation). We used conventional CT images reconstructed with IR (iDose level 3; Philips Healthcare) as controls. We also reconstructed the quantitative, iodine density, and effective atomic number images. The slice thickness of all CT images was 1 mm.

Quantitative image analysis
A radiology technologist with 15 years of experience with cardiac CT performed the quantitative analysis of the axial images. CT attenuation of normal myocardium (HU normal ) and hypo-perfused myocardium (HU hypo ) was measured by placing circular region of interests in axial cardiac CT images. Normal and hypo-perfused areas were determined by invasive coronary angiography. We attempted to select a region of interest of 100 mm 2 in the myocardium that excluded the vessels and perivascular fat. The image noise was de ned as the standard deviation (SD) of the attenuation of the normal myocardium. The contrast and contrast-to-noise ratio (CNR) were calculated with the following formula: We de ned the optimized energy level as that which results in images with the highest CNR. We also measured the iodine concentration and effective Z in the normal and hypo-perfused areas.

Quantitative image analysis
The image quality obtained with the different sequences was evaluated by qualitative image analysis with a PACS viewer (View R, version 1.09.15, Yokogawa Electronic, Tokyo, Japan). Two board-certi ed radiologists with 20 and 13 years of experience with cardiac CT, respectively, independently graded the image contrast, noise, artifacts, sharpness, and overall image quality.
The CT datasets were randomized, and the readers were blinded to the acquisition parameters. Using a subjective four-point scale, they independently graded image contrast and overall quality (1 = unacceptable, 2 = acceptable, 3 = good, or 4 = excellent). Image noise and artifacts were similarly recorded as grade 1 (present and unacceptable), 2 (present and interfering with the depiction of adjacent structures), 3 (present without interfering with the depiction of adjacent structures), or 4 (no noise or artifacts). Image sharpness was determined by evaluating the aortic wall sharpness as grade 1 (blurry), 2 (poorer than average), 3 (better than average), and 4 (sharp). Any disagreement between the readers was settled by consensus.

Statistical analysis
We performed statistical analyses with the programing software Python (version 3.6.3). The Kolmogorov-Smirnov test was performed to determine the normality of the distributions. All numerical values were expressed as mean ± SD. For quantitative image analysis, we performed the paired t-test to compare the best CNR VMIs and conventional CT images and to compare the iodine concentration and the effective Z in the normal and hypo-perfused areas. For qualitative image analysis, we performed the Wilcoxon Signed Ranks test to compare the best CNR VMIs and conventional CT images. Furthermore, the degree of agreement between two observers regarding the visual evaluation results was measured using kappa statistics: poor (<0.20), fair (0.21-0.40), moderate (0.41-0.60), substantial (0.61-0.80), and near-perfect (0.81-1.00). Values of p < 0.05 were considered statistically signi cant.

Results
Quantitative image analysis Figure 1 shows the CT number, image noise, contrast, and CNR between the normal and hypo-perfused myocardium. As the X-ray energy decreased, venous attenuation, contrast, and CNR gradually increased. The 40 keV image offered the best CNR, and thus 40 keV images were selected as the optimized energy VMIs. Figure 2 shows the qualitative analysis regarding the optimized energy VMIs (40 keV images) and 120 kVp images. There was no signi cant difference in image noise between the 120 kVp images (12.6 ± 4.8 HU) and 40 keV images (16.6 ± 13.0 HU) (p = 0.07). The contrast (120 kVp: 41.5 ± 20.4 HU vs 40 keV: 1113.4 ± 49.0 HU) and the CNR (120 kVp: 3.6 ± 2.0 vs 40 keV: 9.1 ± 6.7) between the normal and hypoperfused myocardium were signi cantly higher in 40 keV images than those in 120 kVp images (p < 0.01). Figure 3 shows the iodine concentration and effective Z in the normal and hypo-perfused areas. The iodine concentration and effective Z were signi cantly higher in normal myocardium (1.4 ± 0.6 mgI/mL and 8.1 ± 0.2, respectively) than those in hypo-perfused myocardium (0.3 ± 0.3 mgI/mL and 7.4 ± 0.4, respectively) (p < 0.01). Table 2 summarizes the qualitative analysis. The 40 keV images offered signi cantly higher scores for the image contrast, noise, artifacts, and overall quality (p < 0.01). Conversely, there were no signi cant differences in the sharpness between the 120 kVp and 40 keV images (p = 0.83). Inter-observer agreements regarding the image contrast, noise, artifacts, sharpness, and overall quality were moderate (kappa = 0.71, 0.69, 0.70, 0.65, and 0.65, respectively). Representative cases are shown in Figure 4 and 5.

Discussion
Our study suggests that the best CNR VMIs between the normal and hypo-perfused myocardium are obtained with 40 keV images. The image noise of 40 keV images was slightly (but not signi cantly) higher than that of conventional CT images. Conversely, the contrast and CNR of 40 keV images were signi cantly higher than those of conventional CT images.
To date, some studies have evaluated CT myocardial perfusion in potential ACS patients [16][17][18][19]. Using a 16-MDCT protocol for 69 patients presenting with acute chest pain, Lessick et al. [16] demonstrated that CT-MPI hypo-perfusion had a sensitivity of 67% and a speci city of 95% for diagnosing myocardial ischemia or infarction. However, most studies acquired myocardial perfusion images using continuous shooting scan that requires additional radiation doses and injections of contrast material. Nagao et al.
reported that there was no signi cant difference in diastolic perfusion between ischemic and nonischemic segments in a conventional CT scanning [19].
In this study, the image noise of VMIs at all energy levels was not signi cantly higher than that of conventional CT images, and the image contrast and CNR between the normal-and hypo-perfused myocardium were highest with 40 keV. The DL-DECT system simultaneously collects low-and highenergy data in the two detector layers at the exact same anatomical location. Therefore, the VMI reconstruction technique can suppress beam-hardening artifacts and anti-correlated noise in photoelectric and Compton scatter images. In this study, the artifact score at 40 keV was lower in VMIs than that in conventional CT images; however, this difference was not statistically signi cant. Modelbased IR techniques equipped with DL-DECT can achieve further noise reduction [20]. In our study, the IR technique decreased the image noise of VMIs compared with that of conventional CT images.
In addition, we found that the iodine density and effective Z in the ACS regions were signi cantly lower than those in the normal regions. Although the accuracy of this technique will need to be compared with other techniques, such as myocardial scintigraphy and CT perfusion, we believe that this technique has the potential to quantitatively assess hypo-perfusion in ACS.
Our study has some limitations. First, the study population was small, and our results might not be generalizable. Second, we did not perform diagnostic performance analysis of culprit vessels because of the limited number of patients with multi-vessel stenoses.
In conclusion, spectral imaging with dual-layer spectral detector CT is a feasible technique to detect hypoperfused areas of acute ischemic syndrome.

Declarations
Funding: Not applicable.
Con icts of interest/Competing interests: The authors declare no con icts of interest associated with this manuscript.
Availability of data and material: Not applicable.
Code availability: Not applicable.

Authors' contributions:
Ethics approval / Consent to participate / Consent for publication: This retrospective study received institutional review board approval; the requirement for written informed consent was waived.    Quantitative image analysis The computed tomography numbers (a) of 40 keV images are signi cantly higher than those of 120 kVp images of the normal myocardium, whereas the increase is not so notable for hypo-perfused myocardium. The image contrast (b) is signi cantly higher in 40 keV images than that in 120 kVp images. The image noise (c) does not signi cantly differ between the two protocols. The contrast-to-noise ratio (d) is signi cantly higher in 40 keV images than that in 120 kVp images.  Case 1 A 75-year-old obese man with no prior cardiac history presented with chest pain. Coronary artery evaluation showed severe stenosis in LAD and D2. However, it was not possible to determine the culprit lesion. We performed a myocardial perfusion evaluation. Compared with 120 kVp (A), the 40 keV image (B) clearly depicts the hypo-perfusion area in the D2 segment. The iodine density map (C) and effective Z map (D) clearly depict the iodine defect in the D2 segment. These results suggest that the culprit vessel is