Potential of optical frequency domain imaging for differentiation between early and advanced coronary atherosclerosis

This study evaluated whether optical frequency domain imaging (OFDI) accurately distinguish between fibroatheroma (FA) and pathological intimal thickening (PIT) compared with histopathology. A total of 631 histological cross-sections from 14 autopsy hearts were analyzed for the comparison between OFDI and histological images. Of those, 190 (30%) sections were diagnosed with PIT and 120 (19%) with FA. The OFDI signal attenuation rate was calculated from an exponential. The lipid length was measured longitudinally by detection of sequential OFDI frames within a plaque segment containing lipids. The lipid arc was measured with a protractor centered in the center of the lumen. The fibrous cap thickness was defined as the minimum thickness of the signal rich band overlying PIT and FA. There was no significant difference in the OFDI signal attenuation rate between FA and PIT (3.09 ± 1.04 versus 2.79 ± 1.20, p = 0.13). However, the lipid length was significantly longer, the maximum lipid arc was significantly larger, and the fibrous cap thickness was significantly thinner in FA than in PIT (7.5 [4.3–10.3] mm versus 4.3 [2.7–5.8] mm, p < 0.0001, 125 [101–174]° versus 96 [74–131]°, p < 0.0001, and 220 [167–280] µm versus 260 [190–332] µm, p = 0.019). This study revealed OFDI may have the potential capability for discriminating FA from PIT based on the longitudinal and circumferential extent of lipid plaque, although the OFDI signal attenuation rate was similar between FA and PIT.


Introduction
Over the decades, atherosclerosis of human coronary artery progresses in stages, with early atherosclerotic lesion developing into advanced stage lesion, under the influence of various factors. Pathologic intimal thickening (PIT) is an early atherosclerotic lesion consisting of proteoglycans and that lipid components are common to both FA and PIT, it is unclear whether OCT can be used to accurately distinguish between FA and PIT. Therefore, in the current study, we analyzed the differences in OFDI findings between FA and PIT in comparison with histopathology.

Study subject
Thirty-nine coronary arteries from 14 human cadaver hearts were analyzed for the comparison between OFDI and histological images. All three major coronary arteries were collected from 11 patients, whereas only two major coronary arteries were collected from three patients due to hypoplasia of the coronary arteries. The cause of death was cardiovascular disease in seven cadavers, non-cardiac-related in six cadavers, and unknown in the remaining one cadaver. The experimental study protocol was approved by the Institutional Review Board of Hyogo College of Medicine (approval number 3766).

OFDI imaging procedure
Three major coronary arteries, including the left anterior descending artery, the circumflex artery, and the right coronary artery with the surrounding fatty tissue, were carefully removed from the autopsy hearts after death for ex vivo OFDI imaging. The surrounding fatty tissue was carefully dissected from each coronary specimen. Before OFDI examination, using a tapered surgical needle, multiple 6 − 0 proline sutures were carefully inserted into the plaque segment as a reference point for matching between the OFDI and histological images. This method was successfully performed in previous comparative studies that compared histological findings and intravascular ultrasound (IVUS) images 6,7. Side branches were tied off with sutures to preserve a perfusion pressure of main vessels during the OFDI examination. A 0.014-inch guidewire was advanced to the distal end of each harvested major coronary artery, followed by an OFDI catheter (LUNAWAVE, Terumo Corporation, Tokyo, Japan). OFDI images of the entire vessel were acquired at a pullback rate of 20 mm/s (160 frames/s).

Histopathological preparation and assessment
After OFDI examination, each coronary artery was fixed in 10% neutral buffered formalin for 48 h. The ring-like arterial specimens obtained at the same level as the imaging study were decalcified for 5 h, before being embedded in paraffin and cut every 3 mm into 4-µm transverse sections perpendicular to the longitudinal axis of the artery. All histological sections were stained with hematoxylin & eosin, elastic van Gieson, and Masson's trichrome stain. Coronary plaques were classified using the modified American Heart Association classification 2 into the following categories: Adaptive intimal thickening, defined as lesions with predominantly fibrous tissue and no macrophage and lipid pool; PIT, defined as lesions with lipid pools and no apparent necrosis; FA, defined as lesions comprised of lipid pools and necrotic cores, including cholesterol clefts and cellular debris with an overlying fibrous cap; fibrocalcific plaque, defined as lesions with calcification and absence or fractions of a necrotic core; and healed plaque, defined as lesions composed of smooth muscle cells, proteoglycans, and collagen type III, with or without an underlying disrupted fibrous cap and necrotic core. Histological assessment of each crosssection was performed by a single experienced pathologist (R.K.) who was blinded to the OFDI findings. When the pathologist (R.K.) had difficulty in making a pathological diagnosis, the final diagnosis was determined by discussing with another pathologist (H.H.).

Co-registration of OFDI images with histology
All OFDI images were co-registered with histologic sections by an experienced investigator. Adjustments were made using the sutures and luminal configuration or anatomical landmarks such as vessel branches, thus improving the accuracy of registration. A total of 228 histological segments were excluded due to difficulties with co-registration. Finally, 631 pairs of matched images were acquired from OFDI with corresponding histological sections.

Quantitative OFDI analysis
OFDI image analysis was performed only when histological cross-sections of paired images were classified as a FA or PIT. An OFDI-derived lipid image was defined as a diffusely bordered, signal-poor region 8, 9. OFDI was reviewed longitudinally, and the lipid length was measured by detection of sequential OFDI frames within a plaque segment containing lipids. The lipid arc measurements were made by defining the circumferential extent of the lipidic tissue from the lumen center at the cross-section of minimum lumen area. The fibrous cap thickness was defined as the minimum thickness of the signal rich band overlying the OFDI-derived lipid image (Fig. 1). The OFDI signal intensity was calculated on each corresponding OFDI cross-section using Image J software (National Institutes of Health, Rockville, Maryland, USA). Because the OFDI signal intensity of lipid tissue should gradually decrease from the surface to the inside, the attenuation of lipid signal intensity was analyzed by fitting it to a single exponential function: y = A × exp − Bx , where index B represents an "attenuation rate" (Fig. 2). OFDI images were analyzed by an experienced observer (T.I.) who was blinded to the clinical and histopathological presentations.

Statistical analysis
Continuous variables are expressed as mean ± standard deviation or median (Q1-Q3). The normality of the distribution was analyzed using the Shapiro-Wilk test, and homoscedasticity was analyzed using the Bartlett test. Continuous variables were compared using Student's t-test or Wilcoxon rank sum test. The area under the receiver operating characteristic (ROC) curve was evaluated to determine the best cutoff of lipid length and the FA angle. P-values < 0.05 were considered statistically significant. Statistical analyses were performed with the use of JMP Pro 14.2.0 (SAS Institute Inc., Cary, NC, USA).

Quantitative OFDI analysis
All cross sectional images of both PIT and FA showed low intensity area with diffuse border. There was no significant difference in the optical attenuation coefficient between FA and PIT (3.09 ± 1.04 versus 2.79 ± 1.20, p = 0.13) (Fig. 3 A). However, the lipid length was significantly longer in FA than in PIT (7.5 [4.3-10.3] mm versus 4.3 [2.7-5.8] mm, p < 0.0001) (Fig. 3B). ROC analysis identified 5.8 mm as the optimal cutoff point for prediction of the lipid length of FA (sensitivity, 66% and specificity, 74%; area under the curve, 0.73; p < 0.0001) (Fig. 4 A). Furthermore, the maximum lipid arc was significantly larger in FA than in PIT (125 [101-174]° versus 96 [74-131]°, p < 0.0001) (Fig. 3 C). ROC analysis identified 113° as the optimal cutoff point for prediction of the maximum FA angle (sensitivity, 70% and specificity, 68%; area under the curve, 0.71; p < 0.0001) (Fig. 4B). The fibrous cap thickness was significantly thinner in FA compared with PIT (220 [167-280] µm versus  The OFDI signal intensity was calculated from the luminal surface by Image J software, and fitted to the following approximate formula: y = A × exp − Bx . Index B reflects an "attenuation rate". This index rate was defined as an optical attenuation coefficient. The OFDI signal attenuation rate measured from the luminal surface to the leading edge of the outer lumen was 4.23 OFDI = optical frequency domain imaging OFDI is accepted as a high-resolution intracoronary imaging modality to assess morphological features of native coronary arteries in comparison with other imaging modalities 10, 11. Given that it is recognized that OFDI is capable of differentiating lipid tissue from fibrous and calcific tissue, OFDI is widely used as a promising imaging modality for identifying vulnerable plaques in clinical settings. Pathologic features of vulnerable plaques are characterized by a large necrotic core with an overlying thinner fibrous cap 1, 12. Moreover, on OFDI, lipidic tissues appear as signal-poor regions with diffuse borders because of multiple scattering in lipids at light wavelengths of approximately 1,300 nm 13. A previous ex vivo imaging study validated this finding with ex vivo OCT imaging of diseased atherosclerotic arterial segments obtained at autopsy, and reported a sensitivity and specificity of 90-94% and 90-92% for lipidic plaques, respectively 5. This finding was confirmed 260 [190-332] µm, p = 0.019) (Fig. 3D). ROC analysis identified 240 μm as the optimal cutoff point for prediction of the fibrous cap thickness (sensitivity, 66% and specificity, 58%; area under the curve, 0.61; p < 0.038) (Fig. 4 C). Representative images of FA and PIT are shown in Fig. 5.

Discussion
The main findings of this study are as follows: (1) the lipid length was significantly longer and the maximum lipid arc was significantly larger in FA than in PIT, and (2) the OFDI signal attenuation rate was similar between FA and PIT. To the best of our knowledge, this is the first study to report the differences in OFDI findings between "lipids" at an early stage and those at an advanced stage, in comparison with histopathology. characterizing thin-cap FA, using histology as a standard, was only 41% 15 . The majority of tissue, falsely diagnosed as thin-cap fibroatheroma on OCT, was categorized as PIT. In line with our findings, a previous study showed that CD68stained dense foam cell infiltration creates a highly scattered layer that casts a dark shadow on the tissue behind, and therefore appears as a thin-cap FA 15. In the current study, PIT without a necrotic core, which was considered by another ex vivo study that showed > 90% sensitivity and specificity for identifying lipidic plaques by OCT in 40 human cadavers 14. However, there is a pitfall in the diagnosis of lipidic plaques by OFDI in that the OFDI signal could be attenuated because of multiple scattering, not only in the necrotic core, but also in any lipid components, because of the features of near-infrared light. Therefore, we previously reported that the diagnostic accuracy of OCT for  larger amount of lipid components than PIT. Although several studies have reported the usefulness of embolic protection devices to reduce the risk of intraprocedural distal embolization, previous prospective randomized trials failed to show the effectiveness of its routine use during the procedure 26−28 . Fujino et al. 29 analyzed pathological characteristics of tissue captured using by the filter-based embolic protection devise during percutaneous coronary intervention (PCI) in the patients with stable coronary artery disease and ACS. They found that the presence of necrotic debris within plaque was a higher risk for distal embolism during coronary intervention. Moreover, Hibi et al. 30 demonstrated that the selective use of the distal embolic protection device during stenting of lesions with attenuated plaque ≥ 5 mm assessed by IVUS was associated with a lower incidence of slow flow compared with stenting without distal protection. These data may support the hypothesis that specific plaque morphologies are more prone to atheroembolism during PCI.
IVUS represents a reasonable option to characterize plaque composition and detect vulnerable plaques that are typically characterized by an eccentric pattern, positive remodeling, and ultrasound attenuation 21-23. Furthermore, near infrared spectroscopy IVUS (NIRS IVUS) has been introduced recently as an intracoronary imaging modality to detect lipid core plaque by spectroscopy and morphological features by greyscale IVUS 24, 25. The use of a hybrid imaging catheter combining OFDI with IVUS or NIRS-IVUS may further improve the characterization of atherosclerotic plaques.
Given the difficulty in differentiating FA from PIT, OFDI observers could misclassify PIT as a vulnerable plaque by only looking at a single cross-section of the OFDI image. The misclassification of plaque morphology in the catheterization laboratory may affect the selected treatment strategy, with the adoption of either pharmacologic 26 or interventional solutions 27. Therefore, interventional cardiologists who are responsible for the interpretation of OFDI imaging should evaluate plaque morphology by looking at the entire segment, not just a single cross-section.

Limitation
There were several limitations in the current study. First, a lack of cardiac motion may have affected the OFDI images acquired ex vivo. Second, a positional discrepancy both OFDI images and histological cross-sections may have influenced the results. Therefore, utmost care was taken to ensure that OFDI images matched the histology. Finally, immunochemical staining to identify macrophages was not performed in the present study. as early-stage atherosclerosis, appeared as a low-signalintensity tissue with diffusely delineated borders due to strong signal attenuation on OFDI. When the tissue contained a lipid component, the attenuation coefficient of the OFDI light signal was similar regardless of the presence of a necrotic core. These results are supported by a previous ex vivo OCT that showed a similar index of plaque attenuation of OCT signal between FA and PIT 16. Because the necrotic core and other lipid components have similar OFDI attenuation coefficients, it remains difficult to distinguish between PIT and FA on a single cross-sectional OFDI image in vivo.
IVUS is also a useful device to characterize plaque composition and detect vulnerable plaques with eccentric pattern, positive remodeling, and ultrasound attenuation 17-19. As mentioned above, we have previously analyzed the diagnostic accuracy of OCT and IVUS for detecting thincap FA of human autopsy specimens, with histology as the gold-standard 20. Although the diagnostic accuracy of OCT and IVUS for detecting thin-cap FA were 41% and 19%, respectively, the combined use of OCT and IVUS increased the diagnostic accuracy up to 69%. Therefore, the use of a hybrid imaging catheter combining OFDI and IVUS may provide more accurate diagnosis of atherosclerotic plaque in the clinical catheterization laboratory. First, the entire lesion should be reviewed by IVUS, and when there is a plaque with positive remodeling and strong ultrasound attenuation, it should be observed by OFDI. If OFDI showed larger arc and longitudinally longer low intensity area with thin fibrous cap observed, it could be considered as vulnerable FA. Furthermore, near infrared spectroscopy IVUS (NIRS-IVUS) has been introduced recently as an intracoronary imaging modality to detect lipid core plaque by spectroscopy and morphological features by greyscale IVUS 21, 22. In this regard, an intravascular combined catheter with dual NIRS-IVUS and OFDI capabilities may provide much more accurate diagnosis of vulnerable FA in vivo.
In the current study, we found that FA had a significantly longer and larger arc of low signal intensity area on OFDI compared to PIT. To date, there are no histopathological data investigating the size of lipid components in FA and PIT. However, PIT is considered a progressive lesion in the early stages of atherosclerosis and represents a precursor lesion to FA 1, 17. Therefore, if PIT is a precursor lesion of FA, it is reasonable to assume that FA contains more lipid components than PIT. A previous clinical study has reported that the length of the lipid pool estimated by OCT was significantly longer in lesions with ≥ 50% ST resolution than in lesions with < 50% ST resolution 18. Furthermore, another OCT study revealed a linear relationship between the intrastent protrusion area and the lipid index, which was calculated as the lipid arc multiplied by the lipid length 19. These clinical data may support our finding that FA contained a