The estimation of coronary artery calcium thickness by computed tomography angiography based on optical coherence tomography measurements

Optical coherence tomography (OCT) is recommended to be the most appropriate modality in assessing calcium thickness, however, it has limitations associated with infrared attenuation. Although coronary computed tomography angiography (CCTA) detects calcification, it has low resolution and hence not recommended to measure the calcium size. The aim of this study was to devise a simple algorithm to estimate calcium thickness based on the CCTA image. A total of 68 patients who had CCTA for suspected coronary artery disease and subsequently went on to have OCT were included in the study. 238 lesions of them divided into derivation and validation dataset at 2:1 ratio (47 patients with 159 lesions and 21 with 79, respectively) were analyzed. A new method was developed to estimate calcium thickness from the maximum CT density within the calcification and compared with calcium thickness measured by OCT. Maximum Calcium density and measured calcium-border CT density had a good correlation with a linear equation of y = 0.58x + 201 (r = 0.892, 95% CI 0.855–0.919, p < 0.001). The estimated calcium thickness derived from this equation showed strong agreement with measured calcium thickness in validation and derivation dataset (r2 = 0.481 and 0.527, 95% CI 0.609–0.842 and 0.497–0.782, p < 0.001 in both, respectively), more accurate than the estimation by full width at half maximum and inflection point method. In conclusion, this novel method provided the estimation of calcium thickness more accurately than conventional methods.


Introduction
Coronary calcification may result in inadequate lesion preparation, limit stent delivery and deployment and may result in stent under expansion, strut malapposition, and direct damage to the stent surface (including polymer), with potential impairment of drug delivery even in current percutaneous coronary intervention (PCI). Stent under expansion has shown to be a major determinant of stent thrombosis and restenosis, which carries adverse prognosis and are often difficult to treat. The thickness of calcium predicts whether the cracks can be created by balloon dilation [1,2]. The calcium thickness, angle, length, presence of calcified nodule and vessel diameter has been previously reported as predictors of stent under expansion [3][4][5]. For heavy calcifications with these factors, lesion preparation with atherectomy or lithotripsy devices is recommended [4], so assessing the thickness of the calcification in advance is an important element in determining the treatment strategy. Although, intracoronary imaging modalities are recommended to evaluate coronary calcification [4], these are not adequate in terms of intra-operative assessment and limitation of measurable calcification thickness [6,7]. The non-invasive coronary computed tomography angiography (CCTA) has the advantage of high sensitivity for detection of calcium, and coronary artery calcium score has good correlation of the presence of coronary artery disease [8,9]. It has not only the advantage of detectability and pre-operative evaluation but also the disadvantage of insufficient image quality for evaluation [10]. Although recent computed tomography (CT) equipment has considerably cleared the obstacles related to imaging conditions such as tachycardia, arrhythmias and scanning time, the problems of low resolution and artifacts have not been fully solved. In this study, we have compared and correlated the calcification identified on CCTA and optical coherence tomography (OCT). In addition, we have devised a simple equation by comparing calcium information obtained from current CT equipment with the original calcium morphology to investigate whether the problems of low resolution and blooming artifacts can be reduced.

Study population
This was a single-center, retrospective, observational study that enrolled consecutive patients who were suspected to have de novo coronary artery disease and underwent contrast-enhanced CCTA followed by OCT examination before PCI procedure between January 2017 and April 2019 in New Tokyo Hospital. Following patients were excluded; patients with > 6-month period from CCTA to OCT examination, poor CCTA or OCT image, no calcium findings in CCTA or OCT, CCTA scanning with other voltages than 120 kV, too thick calcium to detect outer border by OCT, balloon dilatation or any other angioplasty before OCT observation and unable to identify the location of calcified lesions between CCTA and OCT. All study procedures were carried out in accordance with the institutional and national ethical guidelines for human studies. All of the patients gave informed consent to participation in this study. The study protocol was approved by the local ethics committee.

OCT and CCTA protocol
OCT was scanned with a Dragonfly OPTIS or Dragonfly OpStar Imaging Catheter (Abbott, Santa Clara, CA, USA) and ILUMIEN OCT Imaging System. Scanned images were assessed with the OPTIS Mobile System or OPTIS Integrated System (Abbott, Santa Clara, CA, USA). CCTA examination was performed using Brilliance 64 channel-DS (PHILIPS, Buckinghamshire, UK) or SOMATOM Definition AS + (SEMENS, Munich, Germany) both with reconstruction algorithms of Filtered back projection. CT scan detector was 0.625 mm × 64-channels, rotation speed was 420 ms/rotation of Brilliance and 0.625 mm × 64-channels, rotation speed was 300 ms/rotation of SOMATOM. The image was scanned with breath-holding, electrocardiogram (ECG) gating and 120 kV of tube voltage. Iopamidol (Iopamiron 370, Bayer Yakuhin, Ltd, Osaka, Japan) or iomeprol (Iomeron 350, Eisai Co., Ltd., Tokyo, Japan) was injected for the contrast-enhanced scan. Scanning was performed with bolus tracking protocol. Injection volume was determined according to body weight (body weight (kg) × 1.0 mL) and the maximum dose was 75 mL. For pre-medication, 20 or 40 mg of metoprolol was administrated orally 1 h before scanning and 12.5 mg of landiolol (Corebeta, Ono Pharmaceutical Co., Ltd., Osaka, Japan) was injected intravenously according to physician's decision. The CCTA image was analyzed with Aquarius iNtuition (TeraRecon, Inc, Tokyo, Japan) with the slab maximum intensity projection (MIP) method [10][11][12]. Longitudinal image was evaluated with thickness setting of 5 mm and cross-sectional image was with minimum in slab MIP method.

OCT and CCTA image analysis
The thickness of major calcification was manually measured across the visually thickest point in cross-sectional OCT image. The measured segments of calcification were adopted to be morphologically easy to identify. Multiple data were collected if there are multiple calcification which is candidate to measure the thickness. OCT images were analyzed by an experienced investigator who did not know CCTA measured data. CT density profile curve of target calcification was measured at same line as measured in OCT. By contrasting coronary morphology in coronary angiography and CCTA and side-branch distribution and calcification morphology in CCTA and OCT, the target calcium locations were matched in CCTA and OCT. The maximum CT density within the calcification (max Calc density) was collected from CT density profile curve (Fig. 1). Full width at half maximum (FWHM) of the calcium was derived from CT density profile curve (Fig. 2) and inflection point (IP) was from the first derivative curve of CT density profile curve (Fig. 3). CCTA images were analyzed by an experienced investigator who did not know OCT measurement data other than the measurement site.

Estimation of calcium thickness
In the first step ( Fig. 1), the calcium-border CT density was measured by adjusting the measured calcium thickness by OCT (M-Calc-thick) to the CT density profile curve created by CCTA at the same cross-section of calcification. The measured calcium-border CT density (M-Calc-border) corresponding to the M-Calc-thick was measured and their variability was examined. In the second step, the degree of correlation between max Calc density and M-Calc-thick was examined. In the third step, the degree of correlation between M-Calc-thick and M-Calc-border density was examined. In the final step, we attempted to devise a method to estimate the calcium-border CT density from the above correlations. The estimated calcium thickness (E-Calc-thick) will be derived adjusting the estimated calcium-border CT density (E-Calc-border) to the CT density profile curve of calcification.

Statistical analysis
Continuous variables are expressed as mean ± standard deviation when normally distributed, or as median (interquartile range [IQR]) when non-normally distributed. An unpaired t test was used for comparison of continuous variables between the two groups. Comparison of categorical variables between groups was performed by Chi square test. In all analyses, p < 0.05 was accepted as statistically significant. Pearson's correlation coefficient was used to assess correlation between variables. Bland-Altman analyses was used to assess agreement between E-Calc-thick and M-Calc-thick. Interobserver agreement and intraobserver reproducibility of continuous variable measurements were assessed by intraclass correlation coefficients (ICC), using 30 randomly selected CCTA images for assessment of max Calc density and 30 randomly selected OCT images for assessment of M-Calcthick by 2 independent observers and by comparing initial and 1-month after analysis of one observer. All analyses were performed with SPSS version 25.0 (SPSS Inc., Chicago, Illinois). is determined by referring to CCTA and OCT image. CT density profile curve is created by drawing a line across the calcification whose thickness is to be measured. In this case, the max calc density was 1613 HU (C) and the M-Calc-thick by OCT was 1.09 mm (B). The M-Calc-border was derived as 1136 HU with adjusting the M-Calc-thick of 1.1 mm to CT profile curve (C). CCTA coronary computed tomography angiography, max calc density maximum CT density within the calcification, M-Calc-border measured calcium-border CT density, M-Calc-thick measured calcium thickness by optical coherence tomography, OCT optical coherence tomography

Patient and lesion characteristics
During the study period, a total of 152 patients were evaluated with CCTA followed by OCT. 84 patients were excluded as per the criteria mentioned in the methods section ( Fig. 4). A total of 68 patients (238-calcified lesions) with a mean age of 72.6 ± 8.4 (range; 49-86) years were analyzed. The clinical and demographic characteristics are provided in Table 1. Eighty two percent (n = 56) were male, 43% (n = 29) had diabetes mellitus, chronic kidney disease accounted for 6% (n = 4) of the cases, of which 3 were on hemodialysis. All lesions were divided into derivation and validation dataset at 2:1 ratio by block randomization. The derivation datasets were aimed to investigate the correlation between max calc density and M-Calc-border, and to derive an equation to estimate the border CT density of calcification. The validation datasets were aimed to compare the accuracy of the equation with existing border determination methods, FWHM and IP methods. Forty-seven patients with 159 lesions were assigned to the derivation dataset, and 21 patients with 79 lesions were assigned to the validation dataset. 142 of 159 lesions were scanned with PHILIPS CT machine and 17 lesions with SEMENS CT machine in Derivation dataset, 64 of 79 lesions with PHILIPS CT machine and 15 lesions with SEMENS CT machine in Validation dataset. There were no significant differences between two groups except, dyslipidemia was significantly higher in the derivation dataset (89.4% vs. 61.9%, p = 0.016) and hemodialysis patients were more common in validation dataset (0% vs. 14.3%, p = 0.027). Average period from CCTA to OCT examination was 39.8 ± 57.9 days ( Table 1). The calcification of left anterior descending artery was the most common target lesion for observation (Table 2).

CCTA, OCT findings and estimation equation development of calcium thickness
In derivation dataset (Table 3), the max Calc density was 1020 HU [IQR 829-1233] and had good correlation with M-Calc-thick (r = 0.645, p < 0.001) (Fig. 5A). The M-Calc-thick was 0.90 mm [IQR 0.76-1.05] and M-Calcborder, derived by fitting the M-Calc-thick to the CT density profile curve, was 798 HU [IQR 688-929]. These M-Calc-thick and M-Calc-border had a good correlation (r = 0.435, p < 0.0001) (Fig. 5B). Taken together, the above relationships suggest that the max Calc density and M-Calc-border had a good correlation with a linear equation of y = 0.58x + 201 (r = 0.892, 95% CI 0.855-0.919, p < 0.001) (Fig. 6). The E-Calc-border derived by this equation was 793 HU [IQR 682-916] and showed strong Fig. 2 Full width at half maximum method. The border CT density is defined as CT density of half value between the baseline and the peak, and full width at half maximum is defined as the distance between the two border CT density points at both sides agreement with M-Calc-border (r 2 = 0.794, 95% CI 0.732-0.858, p < 0.001) (Fig. 7A). The Bland-Altman plot showed the 95% confidence limits (CL) range of absolute differences between M-Calc-border and E-Calc-border was − 162.0 to 161.7 HU with a mean of − 0.1 HU (Fig. 7B). The E-Calc-thick, derived by fitting the E-Calc-border to the CT density profile curve of calcification, was 0.94 mm [IQR 0.79-1.08] and also showed strong agreement with M-Calc-thick (r 2 = 0.481, 95% CI 0.609-0.842, p < 0.001). The Bland-Altman plot showed the 95% CL of − 0.36 to 0.38 mm with a mean of − 0.01 mm (Fig. 8A).

Interobserver agreement and intraobserver reproducibility
There were good interobserver agreement and intraobserver reproducibility for assessment of max Calc density in CCTA image (ICC = 0.998 and 0.989, respectively) and M-Calc-thick in OCT (ICC = 0.997 and 0.991, respectively). In this verification, the OCT images were re-selected to match the CCTA images and the M-Calcthick was measured.

Discussion
The main findings of this study are: (1) The actual CT density of calcifications are much higher than those previously proposed.
(2) The max Calc density correlates with the calcium thickness. (3) The calcium thickness correlates with the calciumborder CT density. (4) The max Calc density has a good first-order correlation relationship with the calcium-border CT density. Consecutive 152 patients who underwent contrast-enhanced coronary computed tomography angiography (CCTA) followed by optical coherence tomography (OCT) were enrolled. After screening with exclusion criteria, 68-patients (238-calcified lesions) were divided into derivation and validation dataset at 2:1 ratio (47 patients with 159 lesions and 21 with 79) and analyzed Two approaches to determine the border of calcification currently have been attempted: one is to reduce blooming artifacts using various reconstruction algorithms [13][14][15][16][17][18], but this is still insufficient to determine visually, and there are no studies evaluating its accuracy in comparison with in vivo calcium morphology. Although there are some reports that absolute CT density that define the calciumborder correlate with the calcium volume and thickness, the agreement is low due to large variation in the border CT density of each calcification, and this method has limitations [19][20][21]. This is the first study to report to derive and validate an equation for estimating calcium thickness of CCTA images. As a first step, we measured the calcium-border CT density according to the M-Calc-thick, and the median score was 798 HU [IQR 688-929], which is much higher than the 130 HU used in the coronary artery calcium (CAC) score [8] and shows considerable variability. Monizzi et al. reported a good correlation between the area of calcification measured by CCTA and OCT, defining a calcified plaque on CCTA as > 400 HU, but the CCTA measurement was 1.6 times greater than OCT [19]. Therefore, the CT density of actual histological calcification is likely to be much higher than previously recognized. In the second step, we confirmed a good correlation between the max Calc density and M-Calcthick. Okubo et al. also reported that the max Calc density correlated with the thickness and cross-sectional area of the calcification measured by OCT, suggesting that thicker the calcification, higher the max Calc density [20]. In the third step, we confirmed a good correlation between M-Calc-thick and M-Calc-border. The study by Monizzi et al. suggests that as the cross-sectional area of calcification is larger and thicker, calcification measured by CCTA, defined as > 400 HU, tends to be overestimated compared to OCT [19]. Our study, similarly confirmed a positive correlation between calcium size and borderline CT density, suggesting that thicker the calcification, higher the calcium-border CT density. As a final step, we devised a method to estimate the calcium-border CT density from the max Calc density. As mentioned above, the max Calc density correlates with M-Calc-thick, and the M-Calc-thick well correlates with M-Calc-border. By combining these two correlations, we concluded that the max Calc density correlates with M-Calcborder. Since this correlation is close to a linear function, we devised the equation as follows: This E-Calc-border provides E-Calc-thick by fitting the CT density profile curve of calcification (Fig. 10). We evaluated the validation of E-Calc-thick from E-Calc-border, and M-Calc-thick, and found good agreement. Meanwhile, the E-Calc-thick from FWHM had no agreement and from IP less agreement with M-Calc-thick, indicating the superiority of estimation by this equation in this study. It is currently recognized that the calcium CT density is ≥ 130HU, as indicated by the CAC score. However, it has been reported that there is a large variability in calcium CT density [19,22], and the CAC score itself has not been evaluated for the validity of the calcium CT density, but rather as a means to evaluate the prevalence of coronary artery disease. Since the border of tissue exists in the band between the maximum and minimum CT density of the region, FWHM is measured by defining the half value in this band as the border E − Calc − border = 0.58 × (max Calc density) + 201 Fig. 6 Correlation between maximum CT density within the calcification and measured calcium-border CT density. Maximum CT density within the calcification and measured calcium-border CT density had a good correlation with a linear equation of "y = 0.58x + 201"   CT density for convenience [23,24]. Therefore, although FWHM is a simple method, it lacks scientific basis. The calcium CT density is particularly high, which gap between the other surrounding tissue is large, so the error in setting the border CT density is assumed to be large. Therefore, it is difficult to estimate the calcium thickness and morphology by CCTA, and current CT equipment is not recommended for this purpose [25,26]. There have been reports suggesting the possibility of assessing the calcium volume and morphology by CT recently [19][20][21], but no specific calcium-border CT density have been proposed, and clinical usefulness has not yet been established. We speculated that the absolute calcium-border CT density itself does not exist, and that there is variability among calcifications. In this study, the thicker the calcification measured by OCT, the higher the max Calc density tended to be, suggesting that the max Calc density was a good predictor of calcium thickness. Finally, it was found that the max Calc density was the defining factor of the calcium-border CT density, leading to the development of the equation "E-Calc-border = 0.58 × (max Calc density) + 201". The thickness and morphology of the calcification are important predictors in the creation of cracks by the balloon and subsequent stent expansion. Therefore, estimating these factors is important in developing PCI strategies for calcified lesions. It has been reported that the predictors of stent underexpansion for calcified lesions are; maximum calcium angle > 180°, maximum thickness > 0.5 mm, and length > 5 mm by OCT estimation [3], maximum superficial calcium angle > 270° in ≥ 5 mm length, superficial circumferential calcium, calcified nodule by intravascular ultrasound (IVUS) estimation [5], and ablation techniques with Agreement between estimated calcium thickness, from full width at half maximum and inflection point, and measured calcium thickness in validation dataset. Scatter plots of relationship and Bland-Altman plots of the mean differences between E-Calc-thick from FWHM (A) and IP (B), and M-Calc-thick showed poor agreement. E-Calc-thick estimated calcium thickness, FWHM full width at half maximum, IP inflection point, M-Calc-thick measured calcium thickness by optical coherence tomography, SD standard deviation atherectomy devices is recommended to be used aggressively for the lesions with these factors [4]. Intravascular imaging with IVUS or OCT is recommended to assess the thickness and morphology of calcification, however, there are two problems with this evaluation; first of all, the intravascular imaging modalities are limited in their ability to Fig. 10 Simple algorithm to estimate calcium thickness by coronary computed tomography angiography. This study compared and correlated calcium evaluation on OCT and CCTA and devised a simple algorithm to estimate calcium thickness with the equation "E-Calcborder = 0.58 × (max Calc density) + 201" based on the CCTA measurement. In the first step of the novel algorism, a CT density profile curve of target calcification to be measured in cross-sectional CT image is created and the max Calc density is obtained (A, B). In the second step, the E-Calc-border is calculated by applying the max Calc density to the equation "E-Calc-border = 0.58 × (max Calc density) + 201" (B). In the final step, the distance between the two points corresponding to the E-Calc border is measured on the CT density profile curve of the target calcification to obtain the E-Calcthick (C). In the present case, max Calc density = 1228 HU, E-Calcborder = 913 HU, and E-Calc-thick = 1.3 mm, showing high agreement with optical coherence tomography measurements of the same section (D). E-Calc-border estimated calcium-border CT density, E-Calc-thick estimated calcium thickness, max calc density maximum CT density within the calcification depict calcifications. In addition, IVUS cannot mechanistically assess the thickness, and OCT may not be able to visualize the entire calcification depending on the catheter position and calcium size. Finally, all the intravascular imaging modalities are used in the setting of PCI (intraoperative) and cannot be evaluated prior to the procedure (pre-operative) to plan the strategy. However, the CCTA, can visualize the entire calcification and can be evaluated before PCI given the advantage of its non-invasive nature of the examination. These are great advantages compared to intravascular imaging, however, its poor quantitativeness of calcification due to low resolution is big disadvantage. This study is expected to improve this problem and enable the pre-operative assessment of calcification using CCTA and the development of treatment strategies for PCI of calcified lesions.

Study limitations
There are some limitations in our study; first of all, this is a single-center study with a small number of lesions. Second, evaluation potential in different CT systems or reconstruction algorithms were not examined. Third, differences in analysis capabilities between CT workstations were not examined. In addition, most workstations commercially available in Japan are able to create detailed CT profile curves and each parameter can be derived from the CT profile curve, although some workstations are unable. Fourth, different imaging conditions, such as subject's body size, intravascular contrast density, and image quality, has not been evaluated. In addition, the differences of CT density in different locations of the coronary lesion have not been examined. The observation points of CCTA and OCT may not be completely identical. Finally, the calcifications that were too large or too small to measure thickness by OCT were not evaluated.

Conclusions
We verified that the max Calc density is a determinant of calcium thickness. The equation derived from this correlation allow for more accurate calcium quantification than conventional methods. This novel method may make CCTA a useful pre-operative evaluation modality in PCI for calcified lesions in addition to intravascular imaging. Further verification of the correlation with clinical treatment results is required to confirm the clinical usefulness of this method.
Funding The authors have no relevant financial or non-financial interests to disclose.

Data availability
The all supporting data of this study are available from the corresponding author upon reasonable request.