Plaque burden can be assessed using intravascular optical coherence tomography and a dedicated automated processing algorithm: a comparison study with intravascular ultrasound

Abstract

Aims

Plaque burden (PB) measurement using intravascular optical coherence tomography (IVOCT) is currently thought to be inferior to intravascular ultrasound (IVUS). We developed an automated IVOCT image processing algorithm to enhance the external elastic lamina (EEL) contour. Thus, we investigated the accuracies of standard IVOCT and an IVOCT enhancement algorithm for measuring PB using IVUS as the reference standard.

Methods and results

The EEL-enhancement algorithm combined adaptive attenuation compensation, exponentiation, angular registration, and image averaging using three sequential frames. In two different laboratories with intravascular imaging expertise, PB was quantified on 200 randomized, matched IVOCT and IVUS images by four independent observers. Fibroatheroma, fibrocalcific plaque, fibrous plaque, pathological intimal thickening (PIT), and mixed plaque were included in each set. Pearson’s correlation coefficients between IVUS and standard IVOCT measurements of PB were 0.61, 0.67, 0.76, 0.78, and 0.87 for fibroatheromas, mixed plaques, fibrocalcific plaques, fibrous plaques, and PIT plaques, respectively. Pearson’s correlation coefficients increased to 0.81, 0.83, 0.83, 0.84, and 0.90 when using the EEL-enhanced images (P = 0.003, P = 0.004, P = 0.08, P = 0.12, and P = 0.23, respectively). EEL-enhanced IVOCT analysis was associated with a lower EEL-area measurement absolute error for fibroatheromas, mixed plaques, and all pooled plaques (P = 0.006, P = 0.02, and P < 0.001, respectively). Compared with standard IVOCT, the EEL-enhanced IVOCT images had a higher sensitivity (79% vs. 28%, P < 0.001) and specificity (98% vs. 85%, P = 0.03) for plaques with an IVUS PB ≥70%.

Conclusion

EEL-enhanced IVOCT can be used to reliably measure PB in all types of coronary atherosclerotic lesions, including fibroatheromas and mixed plaques.

Introduction

Coronary atheroma progression including positive remodelling is linked to future clinical events such as acute coronary syndrome and sudden cardiac death.^1–3 The extent of atherosclerotic plaque is often quantified as plaque burden (PB), which is the percentage of plaque area within the entire vessel area. The largest prospective intravascular imaging atherosclerosis natural history study (PROSPECT) using intravascular ultrasound (IVUS) found that a PB ≥70% was an independent predictor of major adverse coronary events at 3-year follow-up.⁴ IVUS and intravascular optical coherence tomography (IVOCT) are high-resolution imaging techniques that are increasingly being utilized in interventional cardiology for the investigation and management of coronary artery disease.^5,6 Although the resolution of IVOCT is 10 times superior to that of IVUS, the IVOCT technique is limited by attenuation of the backscattering signal in tissue.⁷ The limited depth of penetration of IVOCT light at the atherosclerotic lesion site results in the loss of external elastic lamina (EEL) visibility in many lesions with large PB.⁸ Key plaque features such as lipid, for example, manifest as a low IVOCT signal; and frequently the EEL is not readily visualized in these lesions. With the EEL contour of the diseased vessel not clearly visible, errors or difficulties in measuring PB are possible.^9,10 To mitigate this issue, some investigators have used plaque free wall¹¹ or lipid arc measurements as a surrogate marker of PB.¹²

A recent study by Teo et al.,¹³ used attenuated-compensated IVOCT to show a significant improvement in EEL visualization in comparison to histology. Here, we developed an EEL enhancement algorithm that incorporates an attenuation compensation step.^13–15 To validate this new algorithm, we tested whether PB measurements using enhanced IVOCT images are superior to those obtained with conventional IVOCT images, by using measurements of PB from matched IVUS images as the reference standard.

Methods

Study population

Forty-two non-consecutive patients scheduled for elective percutaneous coronary intervention (PCI) were enrolled at two centres [Columbia University Medical Center (CUMC), New York, NY, USA and Lahey Clinic Medical Center, Burlington, MA, USA]. Massachusetts General Hospital (MGH-2005P000646), CUMC (Columbia IRB-AAAC3289), and Lahey IRBs (Lahey IC-ID 2005-46) approved the study protocol. The study complied with the Declaration of Helsinki and informed consent was obtained from all patients. Patients presenting with ST-elevation myocardial infarction, haemodynamic instability, renal insufficiency (glomerular filtration rate <50 mL/min), allergy to X-ray contrast, unprotected left main coronary artery disease, venous bypass graft lesions, chronic total occlusions, last remaining vessel, or extremely tortuous vessels were excluded. Supplementary data online, Table S1 depicts baseline characteristics of the enrolled patients. Patients underwent the following procedures in the catheterization laboratory: coronary angiography, PCI of the culprit lesion, and intravascular imaging (IVUS and IVOCT imaging; Supplementary Materials) in random order.

Matched IVUS-IVOCT images sets

The matched IVUS-IVOCT images database was made by two independent interventional cardiologists (A.M. and A.T.) (Supplementary Materials). Among this initial database, images including stent struts or a lumen diameter >4.5 mm were excluded for this study. Five hundred and fifty matched IVUS-IVOCT images were obtained. Then, each IVOCT image was classified using standard IVOCT image interpretation criteria.^6,16 Tissue type for each frame was categorized as: fibroatheroma, fibrous plaque, fibrocalcific plaque, mixed plaque, and pathological intimal thickening (PIT). This plaque type classification was done by an expert IVOCT image reader in 550 IVOCT frames (E.G.) using IVOCT criteria.^6,16 Lipid plaque was identified as a signal-poor region with diffuse borders and an overlying signal-rich layer.⁶ Fibrous plaque was defined as a lesion with homogeneous IVOCT signal and high backscattering.⁶ Fibrocalcific plaque contained IVOCT evidence of fibrous tissue (defined previously) and calcium plates that appeared as signal-poor or heterogeneous regions with sharply delineated borders.⁶ IVOCT heterogeneous lesions including fibroatheroma and calcific tissue were classified as mixed plaques.¹⁷ PIT was defined as lesion that had intimal thickness ≥0.30 mm without clear evidence of lipid plaque. We defined intima thickness of ≥0.30 mm as pathological on the basis of histological studies of coronary arteries in young adults.¹⁸ Sets of 200 IVUS-IVOCT matched images including fibrotheromas (n = 40), fibrous plaques (n = 40), fibrocalcific plaques (n = 40), mixed plaques (n = 40), and PITs (n = 40) were randomly selected from the 550 matched IVUS-IVOCT images by an independent researcher (R.L.) using randomization software on the website https://www.randomizer.org/.

Plaque burden processing algorithm

We created a dedicated optical coherence tomography (OCT) image processing algorithm to better delineate EEL on the IVOCT images, implemented using ImageJ software.¹⁹ The automated processing algorithm combined adaptive attenuation compensation,^13,15 registration, and image averaging of successive frames from IVOCT pullback datasets (Figure 1). For each target frame of the set, substacks of ±1 frame around the centre frame [frame of interest (FOI)] were extracted from the pullback data. Following removal of tick marks using a median filter and correction of z-offset,⁶ image substacks were converted to polar coordinates. Registration was accomplished using a custom correlation-based technique, implemented in the spatial frequency domain. Following stack registration, we applied an attenuation compensation plus exponentiation algorithm as described by Girard et al. and Teo et al.^13,15 Images were then converted back to Cartesian coordinates. For measurements on the algorithm-modified data, the processed FOI was used for manual lumen segmentation and a frame averaged image (n = 3, centred around the FOI) was employed for manual segmentation of the EEL. All images were corrected for an IVOCT compensation refractive index of 1.4. Prior to segmentation, all images were brightness and contrast-balanced using ImageJ’s integrated contrast normalization function.

Figure 1.

Schematic of the PB dedicated automated processing algorithm. For each target frame of the set, substacks of ±1 frame around the centre frame (frame of interest, FOI) were extracted from the pullback data. Following correction of z-offset, image substacks were converted to polar coordinates. Registration was accomplished using a correlation-based technique, implemented in the spatial frequency domain. Following stack registration, an attenuation compensation plus exponentiation algorithm was applied. Images were then converted back to Cartesian coordinates. For measurements on the algorithm-modified data, the processed FOI was used for manual lumen segmentation and a frame averaged image (n = 3, centred around the FOI) was employed for manual segmentation of the EEL.

IVUS vs. IVOCT data analysis

To assess quantitative PB measurements with and without the IVOCT EEL-enhancement algorithm vs. IVUS, randomized matched datasets of 200 IVOCT and 200 IVUS intracoronary images were analysed by four independent observers from two different laboratories (E.G and P.C. for the IVUS measurements; H.O and G.B for the IVOCT measurements) who were blinded to other data. Two expert IVUS readers (E.G and P.C.) manually delineated IVUS lumen and EEL contours. At the same time two non-expert IVOCT readers [H.O. (pathologist) and G.B. (technician)] manually delineated IVOCT lumen and EEL contours.

IVUS data analysis

IVUS measurements were made on a stand-alone computer workstation using ImageJ software. All quantitative data were evaluated following published consensus document definitions.⁵ To evaluate the intra-observer variability, one observer (E.G.) repeated the analysis of another set after a one-month washout period.

IVOCT data analysis

Anonymized data were analysed on a stand-alone computer workstation using ImageJ software. The following quantitative data were measured: lumen cross-section area (CSA), EEL CSA, atheroma CSA, and PB. EEL and lumen borders were delineated following published consensus document definitions.^6,9 Observers quantified PB on all standard IVOCT images, even when the EEL was not completely identified on the entire circumference; where the EEL was not seen, the border where the IVOCT signal met noise was delineated (Supplementary data online, Figure S1). Similar measurements were then recorded using images processed by the EEL enhancement algorithm. Examples of the IVOCT images used in this study are shown in Figures 2 and 3. To evaluate the intra-observer variability, one observer (H.O.) repeated the measurements one month later.

Figure 2.

Examples of IVUS-IVOCT matched images with corresponding EEL-enhanced IVOCT images analysis using the dedicated PB algorithm. (A) IVOCT image of a fibroatheroma plaque. (B) Corresponding EEL-enhanced IVOCT image enabling improved visualization of the EEL deep to the plaque. (C) Corresponding IVUS image of this fibroatheroma plaque. (D) IVOCT image of another fibroatheroma plaque. (E) Corresponding EEL-enhanced IVOCT image enabling improved visualization of the EEL deep to the plaque. (F) Corresponding IVUS image of this fibroatheroma plaque. (G) IVOCT image of a mixed plaque. (H) Corresponding EEL-enhanced IVOCT image showing a more clearly delineated EEL. (I) Corresponding IVUS image of this mixed plaque.

Figure 3.

Examples of IVUS-IVOCT matched images with corresponding EEL-enhanced IVOCT images analysis using the dedicated PB algorithm. (A) IVOCT image of a fibrocalcific plaque. (B) Corresponding EEL-enhanced IVOCT showing a greatly improved capability to distinguish the EEL. (C) Corresponding IVUS image of this fibrocalcific plaque. (D) IVOCT image of a fibrous plaque. (E) Corresponding EEL-enhanced IVOCT showing much more of the EEL around the artery’s circumference. (F) Corresponding IVUS image of this fibrous plaque. (G) IVOCT image of a pathological intimal thickening. (H) Corresponding EEL-enhanced IVOCT image. (I) Corresponding IVUS image of the pathological intimal thickening.

Quantitative metrics

The following quantitative data were measured for standard and EEL-enhanced IVOCT and IVUS images: lumen CSA, EEL CSA, atheroma CSA (defined by the EEL CSA minus lumen CSA), PB [calculated as (atheroma CSA/EEL CSA) × 100 (%)].

Statistics

Continuous data are expressed as mean ± standard deviation, or median (interquartile range) when appropriate. Intra-observer and inter-observer variability for IVUS and IVOCT quantitative data were determined as mean (relative) difference (bias) and standard deviations according to the methods of Bland and Altman. The limits of agreement were defined as mean ± 1.96 standard deviation of the absolute difference. The relationships between the IVOCT and IVUS measurements were investigated using a linear regression analysis. A Pearson’s correlation coefficient of 0.70 or higher indicates strong positive relationship; a r value of 0.50 to 0.69 indicates moderate positive relationship, and a r value of 0.30 to 0.49 indicates weak positive relationship. For comparisons within Pearson’s correlation coefficients that are themselves inter-correlated, a z-test was performed according Eid et al. by using an online-calculator.²⁰ For comparisons within the magnitudes of intra-observer and inter-observer measurement differences of lumen CSA, EEL CSA, atheroma CSA, and PB between standard IVOCT and EEL-enhanced IVOCT images, a Wilcoxon signed rank test was performed. For comparisons within the absolute differences of EEL CSA measurements between IVUS and conventional IVOCT analyses vs. between IVUS and EEL-enhanced IVOCT analyses in each type of plaque and in all pooled plaques, a Wilcoxon signed rank test was also performed. Using IVUS PB = 70% as a cut-off value, standard IVOCT and EEL-enhanced IVOCT images sensitivity and specificity were calculated and statistically compared using a χ² test with a Yates’ correction. P-values <0.05 were considered significant. All statistics were calculated using NCSS (NCSS 2001; NCSS Statistical software, Kaysville, Utah) and GraphPad Prism 7 (GraphPad Software, La Jolla, CA, USA), except for the comparison of correlations from dependent samples.

Results

Intra-observer and inter-observer variability for IVUS measurements

Intra-observer variability was very low for IVUS lumen CSA, EEL CSA, atheroma CSA, and PB measurements (−0.05 ± 0.27 mm², 0.16 ± 0.82 mm², −0.11 ± 0.98 mm², −0.06 ± 3.1%, respectively). For inter-observer variability, the mean (standard deviation) differences were likewise low for IVUS lumen CSA, EEL CSA, atheroma CSA, and PB measurements (−0.08 ± 0.43 mm², −0.13 ± 1.09 mm², −0.05 ± 1.22 mm², 0.65 ± 5.9%, respectively).

Intra-observer and inter-observer variability for standard IVOCT and EEL-enhanced IVOCT measurements

Intra-observer and inter-observer variability are shown in Table 1. The magnitude of intra- and inter-observer measurement differences for EEL-enhanced IVOCT was statistically significantly less than the magnitude of intra- and inter-observer measurement differences for standard IVOCT for EEL CSA, atheroma CSA and PB.

Table 1.

Intra-observer and inter-observer variability for EEL-enhanced IVOCT vs. standard IVOCT PB measurements

	Mean difference IVOCT	Mean difference IVOCT + algorithm IVOCT A+	Absolute difference IVOCT	Absolute difference IVOCT + algorithm IVOCT A+	P-value
Intra-observer
Lumen CSA (mm²)	−0.05 ± 0.24	0.14 ± 0.12	0.11 (0.05–0.20)	0.09 (0.04–0.18)	0.17
EEL CSA (mm2)	0.13 ± 1.62	0.12 ± 0.90	0.56 (0.23–1.36)	0.44 (0.16–0.92)	0.009
Atheroma CSA (mm2)	−0.18 ± 1.61	−0.02 ± 0.89	0.54 (0.22–1.33)	0.42 (0.17–0.97)	0.009
Plaque burden (%)	0.97 ± 5.0	−0.65 ± 3.1	2.4 (1.2–5.0)	1.9 (0.8–3.2)	<0.001
Inter-observer
Lumen CSA (mm²)	0.02 ± 0.38	−0.09 ± 0.30	0.12 (0.05–0.25)	0.10 (0.05–0.17)	0.08
EEL CSA (mm2)	−0.10 ± 2.27	−0.38 ± 1.03	0.79 (0.26–1.90)	0.60 (0.24–1.23)	<0.001
Atheroma CSA (mm2)	−0.12 ± 2.24	−0.29 ± 1.04	0.80 (0.30–1.95)	0.60 (0.25–1.20)	<0.001
Plaque burden (%)	−1.22 ± 7.6	−0.63 ± 4.9	3.1 (1.3–6.5)	2.4 (0.9–4.4)	0.001

Values are expressed either as mean ± standard deviation or as median (interquartile range). Mean (relative) differences (bias) and standard deviations were calculated according to the methods of Bland and Altman.

Atheroma CSA was calculated as EEL CSA − lumen CSA. Plaque burden was calculated as (Atheroma CSA/EEL CSA) × 100 (%). P-value indicates the use of a Wilcoxon signed rank to compare IVOCT and EEL-enhanced IVOCT intra- and inter-observer mean differences magnitudes.

CSA, cross-section area; EEL, external elastic lamina; IVUS, intravascular ultrasound; IVOCT, intravascular optical coherence tomography; IVOCT A+, intravascular optical coherence tomography measurements using EEL-enhancement algorithm.

Head-to-head comparison of IVUS and IVOCT measurements with and without automated EEL-enhancement algorithm

Fibroatheroma and mixed plaque

The median PB measurements by IVUS for fibroatheromas and mixed plaques were 63% (first quartile: 53%, third quartile: 75%, interquartile range: 22%) and 56% (first quartile: 46%, third quartile: 72%, interquartile range: 26%), respectively. PB ≥70% was measured by IVUS in 28 plaques (35%). The mean differences ± standard deviation of IVUS and IVOCT measurements with and without automated EEL-enhancement algorithm are presented in Table 2. The absolute value of the difference in EEL CSA measurements was reduced from 3.3 mm² (1.9–6.5) to 1.5 mm² (0.6–3.0) for fibroatheromas (P = 0.006), from 3.2 mm² (1.5–4.5) to 2.5 mm² (0.9–3.5) for mixed plaques (P = 0.02), when the automated EEL enhancement algorithm was used instead of standard IVOCT. Bland–Altman showed a moderate agreement for PB in fibroatheromas (5.0 ± 10.5%) and in mixed plaques (1.4 ± 11.4%) without the automated EEL-enhancement algorithm (Figure 4). Using the algorithm, the differences were lower for PB in fibroatheromas (1.7 ± 7.6%) and mixed plaques (2.4 ± 8.6%) (Figure 5). Without the automated EEL enhancement algorithm, Pearson’s correlation coefficients between IVUS and standard IVOCT measurements for PB quantification were 0.61 and 0.67 for fibroatheromas and mixed plaques, respectively (Table 3). Using EEL-enhanced IVOCT images, Pearson’s correlation coefficients increased to 0.81 and 0.83 for fibroatheromas and mixed plaques indicating a very strong positive relationship between IVUS and EEL-enhanced IVOCT PB quantification (Figure 6). The difference with vs. without the dedicated EEL enhancement algorithm for correlations between IVUS and IVOCT PB quantification was statistically significant for both types of plaque (fibroatheroma, P = 0.003; mixed plaque, P = 0.004). Using IVUS PB = 70% as a cut-off value, the EEL-enhanced IVOCT images also enabled a statistically significant improvement in sensitivity [79% (95% confidence interval (CI): 60–90) vs. 28% (95% CI: 15–47), P < 0.001] and specificity [98% (95% CI: 90–99) vs. 85% (95% CI: 76–93), P = 0.03] (Table 4).

Table 2.

Assessment of IVOCT measurements with and without algorithm in comparison with IVUS

	IVOCT	IVOCT + algorithm IVOCT A+	IVUS	Mean difference IVOCT vs. IVUS	Mean difference IVOCT A+ vs. IVUS
Fibroatheroma
Lumen CSA (mm²)	6.12 ± 3.74	6.23 ± 3.80	6.29 ± 3.59	0.16 ± 0.73	0.05 ± 0.91
EEL CSA (mm2)	14.64 ± 8.35	15.83 ± 7.10	16.96 ± 6.28	2.33 ± 6.10	1.13 ± 4.64
Atheroma CSA (mm2)	8.51 ± 5.36	9.44 ± 4.21	10.68 ± 4.39	2.16 ± 5.87	1.23 ± 4.33
Plaque burden (%)	58.6 ± 10.8	61.9 ± 12.0	63.6 ± 12.7	5.0 ± 10.5	1.7 ± 7.6
Mixed plaque
Lumen CSA (mm²)	5.85 ± 3.84	5.89 ± 3.84	6.07 ± 3.70	0.23 ± 0.65	0.19 ± 0.68
EEL CSA (mm2)	12.70 ± 6.21	14.23 ± 7.32	14.33 ± 7.20	1.63 ± 4.34	0.10 ± 3.58
Atheroma CSA (mm2)	6.85 ± 2.99	8.35 ± 4.54	8.25 ± 5.30	1.41 ± 4.21	−0.09 ± 3.31
Plaque burden (%)	55.4 ± 10.6	59.2 ± 11.5	56.8 ± 15.4	1.4 ± 11.4	−2.4 ± 8.6
Fibrocalcific plaque
Lumen CSA (mm²)	6.36 ± 3.26	6.43 ± 3.24	6.43 ± 3.13	0.07 ± 0.60	0.01 ± 0.57
EEL CSA (mm2)	16.36 ± 5.04	16.87 ± 5.88	17.33 ± 4.80	0.96 ± 3.75	0.46 ± 3.62
Atheroma CSA (mm2)	10.0 ± 3.05	10.44 ± 3.90	10.90 ± 2.89	0.90 ± 3.64	0.46 ± 3.44
Plaque burden (%)	62.3 ± 12.0	62.6 ± 11.7	63.9 ± 11.0	1.6 ± 7.9	1.3 ± 6.7
Fibrous plaque
Lumen CSA (mm²)	6.66 ± 2.96	6.82 ± 3.04	6.86 ± 2.77	0.20 ± 0.52	0.04 ± 0.48
EEL CSA (mm2)	13.55 ± 4.32	14.20 ± 5.03	13.82 ± 3.75	0.27 ± 2.65	−0.38 ± 2.62
Atheroma CSA (mm2)	6.89 ± 2.10	7.38 ± 2.49	6.97 ± 2.21	0.08 ± 2.43	−0.42 ± 2.31
Plaque burden (%)	52.4 ± 9.4	53.0 ± 8.8	51.1 ± 11.8	−1.1 ± 7.5	−1.9 ± 6.6
PIT
Lumen CSA (mm²)	7.76 ± 3.46	7.75 ± 3.30	7.91 ± 3.25	0.14 ± 0.51	0.16 ± 0.35
EEL CSA (mm2)	10.81 ± 4.89	10.88 ± 4.83	10.95 ± 4.92	0.14 ± 0.80	0.07 ± 0.59
Atheroma CSA (mm2)	3.05 ± 1.92	3.13 ± 2.12	3.04 ± 2.17	−0.01 ± 0.62	−0.09 ± 0.50
Plaque burden (%)	27.7 ± 8.5	27.7 ± 9.4	26.4 ± 8.2	−1.4 ± 4.2	−1.4 ± 4.1
All plaques
Lumen CSA (mm²)	6.57 ± 3.51	6.62 ± 3.48	6.71 ± 3.34	0.14 ± 0.60	0.09 ± 0.62
EEL CSA (mm2)	13.62 ± 6.16	14.4 ± 6.39	14.68 ± 5.94	1.06 ± 3.99	0.27 ± 3.31
Atheroma CSA (mm2)	7.05 ± 4.04	7.75 ± 4.36	7.97 ± 4.59	0.92 ± 3.83	0.22 ± 3.10
Plaque burden (%)	51.2 ± 16.1	52.9 ± 16.8	52.4 ± 18.3	1.1 ± 8.9	−0.5 ± 7.0

Values are expressed as mean ± standard deviation. Mean (relative) differences (bias) and standard deviations were calculated according to the methods of Bland and Altman.

CSA, cross-section area; EEL, external elastic lamina; IVOCT, intravascular optical coherence tomography; IVOCT A+, intravascular optical coherence tomography measurements using EEL-enhancement algorithm; IVUS, intravascular ultrasound; PIT, pathological intimal thickening.

Figure 4.

Bland–Altman analysis for measurements of standard IVOCT PB (%) in comparison with IVUS. Middle blue line: mean difference; top and bottom dotted lines: mean +1.96 SD and mean −1.96 SD, respectively.

Figure 5.

Bland–Altman analysis for measurements of EEL-enhanced IVOCT PB (%) in comparison with IVUS. Middle blue line: mean difference; top and bottom dotted lines: mean +1.96 SD and mean −1.96 SD, respectively.

Table 3.

Assessment of IVOCT EEL CSA and PB (%) measurements with and without algorithm in comparison with IVUS

	Correlation		Linear regression			Bland–Altman	Limits of agreementa		Correlation		Linear regression			Bland–Altman	Limits of agreementa		Comparison of correlationsb
	IVOCT vs. IVUS		IVOCT vs. IVUS			IVOCT vs. IVUS	IVOCT vs. IVUS		IVOCT A+ vs. IVUS		IVOCT A+ vs. IVUS			IVOCT A+ vs. IVUS	IVOCT A+ vs. IVUS
	Pearson r	P-value	R 2	Equation of line	P-value	Bias, SD of bias	Lower	Upper	Pearson r	P-value	R 2	Equation of line	P-value	Bias, SD of bias	Lower	Upper	P-value
Fibroatheroma
EEL CSA (mm2)	0.69	<0.0001	0.47	Y = 9.42 + 0.52X	<0.0001	2.33 ± 6.10	−9.6	14.3	0.77	<0.0001	0.59	Y = 6.24 + 0.68X	<0.0001	1.13 ± 4.64	−8.0	10.2	0.045
Plaque burden (%)	0.61	<0.0001	0.38	Y = 21.53 + 0.72X	<0.0001	5.0 ± 10.5	−15.6	25.6	0.81	<0.0001	0.66	Y = 10.36 + 0.86X	<0.0001	1.7 ± 7.6	−13.2	16.6	0.003
Mixed plaque
EEL CSA (mm2)	0.69	<0.0001	0.47	Y = 2.75 + 0.68X	<0.0001	1.63 ± 4.34	−6.9	10.2	0.83	<0.0001	0.68	Y = 1.61 + 0.86X	<0.0001	0.10 ± 3.58	−6.9	7.1	0.01
Plaque burden (%)	0.67	<0.0001	0.45	Y = 2.91 + 0.90X	<0.0001	1.4 ± 11.4	−20.9	23.8	0.83	<0.0001	0.68	Y = −9.13 + 1.11X	<0.0001	−2.4 ± 8.6	−19.3	14.5	0.004
Fibrocalcific plaque
EEL CSA (mm2)	0.71	<0.0001	0.51	Y = 6.25 + 0.68X	<0.0001	0.96 ± 3.75	−6.4	8.3	0.79	<0.0001	0.62	Y = 6.47 + 0.65X	<0.0001	0.46 ± 3.62	−6.6	7.6	0.047
Plaque burden (%)	0.76	<0.0001	0.58	Y = 20.38 + 0.70X	<0.0001	1.6 ± 7.9	−13.9	17.2	0.83	<0.0001	0.68	Y = 15.55 + 0.77X	<0.0001	1.3 ± 6.7	−11.8	14.6	0.08
Fibrous plaque
EEL CSA (mm2)	0.79	<0.0001	0.63	Y = 4.50 + 0.69X	<0.0001	0.27 ± 2.65	−4.9	5.5	0.86	<0.0001	0.74	Y = 4.70 + 0.65X	<0.0001	−0.38 ± 2.62	−5.5	4.7	0.031
Plaque burden (%)	0.78	<0.0001	0.61	Y = 3.57 + 0.91X	<0.0001	−1.1 ± 7.5	−15.7	13.5	0.84	<0.0001	0.70	Y = −8.96 + 1.13X	<0.0001	−1.9 ± 6.6	−14.8	10.9	0.12
PIT
EEL CSA (mm2)	0.99	<0.0001	0.97	Y = 0.23 + 0.99X	<0.0001	0.14 ± 0.80	−1.4	1.7	0.99	<0.0001	0.99	Y = −0.06 + 1.01X	<0.0001	0.07 ± 0.59	−1.1	1.2	0.014
Plaque burden (%)	0.87	<0.0001	0.76	Y = 3.37 + 0.83X	<0.0001	−1.4 ± 4.2	−9.6	6.9	0.90	<0.0001	0.81	Y = 4.99 + 0.77X	<0.0001	−1.4 ± 4.1	−9.5	6.8	0.23
All plaques
EEL CSA (mm2)	0.78	<0.0001	0.61	Y = 4.39 + 0.76X	<0.0001	1.06 ± 3.99	−6.8	8.9	0.86	<0.0001	0.74	Y = 3.18 + 0.81X	<0.0001	0.27 ± 3.31	−6.2	6.8	<0.001
Plaque burden (%)	0.87	<0.0001	0.76	Y = 1.41 + 0.99X	<0.0001	1.1 ± 8.9	−16.3	18.6	0.92	<0.0001	0.85	Y = −0.61 + 1.01X	<0.0001	−0.5 ± 7.0	−14.3	13.2	<0.001

Pearson’s correlation coefficients, linear regression analysis (including R², equation of line, and P-values), Bland–Altman analysis (including bias, SD of bias and limits of agreement) and comparisons within Pearson’s correlation coefficients (correlation of standard IVOCT and IVUS measurements vs. correlation of EEL-enhanced IVOCT and IVUS measurements) are successively presented. Values are expressed as mean ± standard deviation (SD).

^aBland–Altman limits of agreement are defined as mean ± 1.96 SD of absolute difference.

^bA z-test was performed to compare Pearson’s correlation coefficients that are themselves inter-correlated. IVOCT A+ indicates intravascular optical coherence tomography measurements using the EEL-enhancement algorithm.

Figure 6.

Assessment of IVOCT PB measurements in fibroatheromas and in mixed plaques with and without algorithm in comparison with IVUS. Regression analysis for measurements of PB percentages between standard IVOCT and IVUS (left panel), between EEL-enhanced IVOCT and IVUS (right panel), in case of fibroatheroma (upper panels), and mixed plaque (lower panels).

Table 4.

Assessment of fibroatheroma and mixed plaques luminal narrowing with and without algorithm in comparison with IVUS

Fibroatheroma and mixed plaques characteristics (n = 80)	IVUS	Standard IVOCT	IVOCT A+
Number of cases with ≥70% luminal narrowing	28 (35)	16 (20)	23 (28)
Matched cases on IVUS and IVOCT measurements		8	22
Number of cases with <70% luminal narrowing	52 (65)	64 (80)	57 (72)
Matched cases on IVUS and IVOCT measurements		44	51

Values in parentheses are percentages.

IVOCT A+, intravascular optical coherence tomography measurements using EEL-enhancement algorithm; IVUS, intravascular ultrasound.

Fibrocalcific plaque, mixed plaque, and pathological intimal thickening

The absolute value of the difference in EEL CSA measurements decreased not significantly from 2.8 mm² (0.9–4.7) to 2.4 mm² (1.1–3.7) for fibrocalcific plaques (P = 0.44), from 2.2 mm² (1.2–3.1) to 1.9 mm² (1.1–2.9) for fibrous plaques (P = 0.40), when EEL-enhanced IVOCT analysis was used instead of standard IVOCT analysis. Concerning PIT plaques, the absolute values of the difference in EEL CSA measurements with and without algorithm were quite similar: 0.4 mm² (0.2–0.7) vs 0.5 mm² (0.2–0.9), respectively (P = 0.22). Pearson’s correlation coefficients values between IVUS and IVOCT measurements for PB were high in fibrocalcific plaques, in fibrous plaques and in PIT plaques (0.76, 0.78 and 0.87, respectively). In the same way, Pearson’s correlation coefficients values between IVUS and EEL-enhanced IVOCT measurements for PB were high in fibrocalcific plaques, in fibrous plaques and in PIT plaques (0.83, 0.84, and 0.90, respectively) (Figure 7). The difference in Pearson’s coefficients value between correlations with and without the automated EEL enhancement algorithm was not statistically significant in fibrocalcific plaques, in fibrous plaques and in PIT plaques (P = 0.08, P = 0.12, and P = 0.23, respectively).

Figure 7.

Assessment of IVOCT PB measurements in fibrocalcific plaques, in fibrous plaques, and in PIT plaques with and without algorithm in comparison with IVUS. Regression analysis for measurements of PB percentages between standard IVOCT and IVUS (left panel), between EEL-enhanced IVOCT and IVUS (right panel), in case of fibrocalcific plaque (upper panels), fibrous plaque (mid panels), and pathological intimal thickening (lower panels).

All pooled plaques

The absolute value of the difference in EEL CSA measurements was reduced from 2.0 mm² (0.7–3.8) to 1.4 mm² (0.6–3.0) for all pooled plaques (P < 0.001), when EEL-enhanced IVOCT analysis was used instead of standard IVOCT analysis. Pearson’s correlation coefficients value between IVUS and IVOCT measurements for PB increased from 0.87 to 0.92, when EEL-enhanced IVOCT analysis was used instead of standard IVOCT analysis (Supplementary data online, Figure S2). The difference between these Pearson’s correlation coefficients values was statistically significant (P < 0.001). Finally, as seen in the Bland–Altman plots in Figure 5, EEL-enhanced IVOCT measurements for PB have a negligible bias of ±1–2% with respect to IVUS.

Discussion

Our results show that for PIT, fibrous, and fibrocalcific plaque, standard IVOCT PB measurements closely approximate those of IVUS; however, the strong association breaks down for fibroatheromas and mixed plaques. EEL-enhanced IVOCT on the other hand was found to be satisfactory for measuring PB in all plaque types, including fibroatheromas and mixed plaques. This result is notable, as coronary atherosclerosis is frequently comprised of many different and potentially complex plaque morphologies. Importantly, the EEL-enhancement algorithm also significantly improved sensitivity/specificity of IVOCT for detecting lesions with PB ≥70%, using IVUS as the reference standard. This finding is significant as PB ≥70% is considered to be the single most predictive morphologic criterion for preemptively identifying the highest risk lesions in patients.^4,21

Conventional wisdom is that EEL under the lipid-containing lesion is less well identified by IVOCT compared with IVUS.^10–12 Delineating EEL on IVOCT images is required to evaluate important features of coronary atherosclerosis such as PB and arterial remodelling. The IVOCT automated EEL-enhancement algorithm developed here combined several approaches including adaptive attenuation compensation, exponentiation for contrast enhancement. In the human optic nerve head, Girard et al.¹⁵ previously demonstrated that the combined use of compensation and exponentiation resulted in higher interlayer contrasts than those obtained with exponentiation only. In the same way, Teo et al. showed that this attenuation-compensated technique could be used on conventional IVOCT images and improve the detection of the EEL. However, this study was performed ex vivo and the EEL area was not quantified.¹³ An additional component of our algorithm was the processing of 3D IVOCT data which included angular registration and image averaging using adjacent IVOCT frames. Thus, depth penetration enhancement was increased due the merging of three images centred on the slice of interest in order to pool similar characteristics of the different layers and constituents of the plaques.

There is an overwhelming perception that IVUS is the only intravascular imaging technique that is capable of accurately measuring PB in vivo. It is well-known that EEL delineation of IVOCT images is dependent on tissue type. Indeed, near-infrared light used by IVOCT is attenuated more rapidly by lipid than with fibrotic or fibrocalcific tissue.^7,22,23 This characteristic of IVOCT has motivated the development of IVOCT measurements that may be related to PB. Kubo et al.¹² for example developed an approximating algorithm for measuring vessel area in coronary arteries with lipid-rich plaque, compared with IVUS. They observed a good agreement between IVOCT and IVUS measured vessel areas (Pearson’s correlation coefficient value of 0.834). However, the mean difference and limits of agreement increased with an increase in lipid arc in their study. Furthermore, this approximating algorithm had some limitations; it could not be used in in oval or irregular vessels, for instance. The EEL-enhancement technique described here has no such limitations.

In our study, the additional value of a dedicated EEL-enhancement algorithm was predominant in fibroatheromas and in mixed plaques (P = 0.003 and P = 0.004, respectively). While the algorithm reported in this article greatly closes the gap between IVOCT and IVUS for CSA and PB measurements, atheroma CSA is still underestimated by IVOCT in fibroatheromas and less so in mixed plaque, possibly owing to residual penetration depth limitations of IVOCT. While EEL enhancement may not be necessary for measuring PB in fibrocalcific plaques, caution must be taken in the analysis of the results because signal attenuation can occur with IVUS in areas of calcium.²⁴ IVOCT may in fact be superior to IVUS for measuring PB in fibrocalcific plaques as shown in Supplementary data online, Figure S3. This figure shows that IVOCT light penetrates calcium, enabling EEL visualization, whereas EEL visualization is difficult for IVUS due to the attenuation of ultrasound by calcium.

Previously, we showed that IVOCT approximates IVUS PB when the EEL is clearly delineated all around its circumference.¹⁰ In the current study, the IVOCT automated EEL-enhancement algorithm helped to reliably visualize the EEL even when the EEL was not completely identified on the entire circumference. Interventional cardiologists using IVOCT are often faced with the limitation of the attenuation and difficulty to obtain good EEL-based measurements in order to choose stent diameters. Interestingly, an EEL-based OCT-guided PCI strategy has been proposed and tested in the prospective, multicentre randomized ILUMIEN III: OPTIMIZE PCI study.²⁵ Interventional cardiologists were able to delineate EEL >180° in the reference segments in 84% of cases (vs. 83% by IVUS), whereas in the core laboratory, the success rates were 95% and 100%, respectively. The use of the automated EEL-enhancement algorithm described here may facilitate EEL delineation for PCI guidance in the catheterization laboratory and for the broader interventional cardiologist community.

Study limitations

Our study included a small number of images (n = 200; 40 per set). However, the analysis was complete, including a wide range of different types of atherosclerotic plaques. To generate the sets, images with poor quality were excluded, which may have induced selection bias, although the investigators who excluded images did not participate in EEL measurement. As with all studies of this sort, the matching process may have had some imperfections. Furthermore, EEL visualization in echo-attenuated plaques by IVUS is sometimes difficult. Depending on the arc of the attenuated plaques, intra-observer and inter-observer variabilities of IVUS PB measurements may have been slightly impacted. This study was conducted in two highly experienced centres in intravascular imaging in the USA and in Europe. Thus, the results of this study may not be applicable elsewhere. Parameters like vascular tonus, vessel location, vessel with a lumen diameter > 4.5 mm, flushing media and variation of cardiac cycle that could affect differences between IVOCT and IVUS assessments were not incorporated in the analysis. Finally, further research is needed to determine whether this EEL enhancement algorithm can be translated into clinical use. Reproduction of this work in larger studies is therefore merited.

Conclusions

A dedicated automated processing algorithm can be applied to conventional IVOCT images in order to improve EEL contour delineation. EEL-enhanced IVOCT measurements of PB were strongly correlated to IVUS PB measurements for all plaque types, even in fibroatheromas and in mixed plaques. Using this algorithm, IVOCT can be used to reliably measure PB.