Analyzer-based phase-contrast x-ray imaging of carotid plaque microstructure

Abstract

BACKGROUND

Plaque vulnerability depends, in part, on composition. Imaging techniques are needed that can aid the prediction of plaque stability. High-contrast images of soft-tissue structure have been obtained with x-ray phase-contrast (PC) imaging. This research investigates multiple image radiography (MIR), an x-ray PC imaging technique, for evaluation of human carotid artery plaques.

METHODS

Carotid plaques were imaged with ultrasound and subsequently excised and formalin fixed. MIR imaging was performed. By using synchrotron radiation, conventional radiographs were acquired for comparison. Image texture measures were computed for soft-tissue regions of the plaques.

RESULTS

Ultrasound evaluation identified plaques as homogeneous without calcifications. MIR images revealed complex heterogeneous structure with multiple microcalcifications consistent with histology, and possessed more image texture in specific regions than conventional radiographs (P < .05). MIR refraction images allowed imaging of the geometric structure of tissue interfaces within the plaques, while scatter images contained more texture in soft-tissue regions than absorption or refraction images.

CONCLUSIONS

X-ray PC imaging better depicts plaque soft-tissue heterogeneity than ultrasound or conventional radiographs. MIR imaging technique should be investigated further as a viable imaging technique to identify high-risk plaques.

Carotid artery plaque instability is a precursor of acute thromboembolic events that lead to ischemic stroke, which is the third leading cause of death in the United States. Patients who are symptomatic with at least 75% stenosis by diameter are candidates for carotid endarterectomy, but this may not be the best option for asymptomatic patients.¹ Plaque vulnerability depends not only on the degree of stenosis but also on its composition. Specifically, plaques with high lipid content and a large necrotic core are more prone to rupture than those with high levels of fibrosis and calcification. Currently available noninvasive imaging modalities cannot accurately determine plaque composition, particularly in the soft-tissue regions. There is a need for the development of improved noninvasive techniques that can identify plaque composition to identify high-risk plaques.

A number of modalities have been investigated for imaging carotid plaques. B-mode ultrasound imaging provides information of plaque echogenicity and structure. Heterogeneity in the ultrasound signal appears to depend on plaque composition and may relate to overall plaque stability.^2–4 However, echogenicity is often not directly related to histologic features. In addition, plaques can produce different echosignals depending on surgeon positioning of the ultrasound transducer, resulting in poor reproducibility. Magnetic resonance imaging (MRI) also has been investigated for plaque imaging. MRI has shown some potential for identifying soft-tissue regions in a plaque as well as quantifying plaque size,^5,6 but MRI has relatively lower spatial resolution, limiting the microstructural features that can be identified. X-ray–based methods provide high spatial resolution and allow identification of calcified regions within plaques. However, soft-tissue structures appear to be the primary contributors to plaque destabilization, which cannot be detected reliably with conventional absorption-based x-ray radiography.

Recent advances in x-ray phase-contrast (PC) imaging allow x-ray imaging that exploits alternative contrast mechanisms based on the x-ray refractive and scattering properties of tissues. A few studies have been performed using an x-ray interferometric method of x-ray PC for imaging plaques from mice.^7,8 In those studies, images acquired were related to the tissue’s mass density to define plaque microstructure and to identify the relative tissue composition of the plaques.^7,8 Although this showed the potential of x-ray PC imaging for visualizing carotid plaques, x-ray interferometry has a small field of view and is difficult to implement in a laboratory environment owing to dependence on synchrotron radiation.

Multiple-image radiography (MIR) is an analyzer-based x-ray PC imaging method that has shown promise for imaging articular cartilage and breast tumors because of its sensitivity for soft-tissue structures.^9,10 It can be used to image a large field of view and currently is being developed for use with bench-top x-ray sources.^10,11 MIR produces 3 separate images that depict the x-ray absorption, x-ray refraction, and ultra-small-angle x-ray scatter (USAXS) properties of tissue. When implemented in computed tomography x-ray (CT) mode, volumetric images of these 3 complementary tissue properties are produced and can provide a detailed characterization of tissue features and microstructure.

In the work described here, we investigated the potential for improved characterization of soft-tissue structures in carotid plaques using MIR. Specific information was identified that can be extracted from each of the 3 MIR images. Image texture measurements of the 3 MIR images and radiographs taken with synchrotron radiation were compared for their ability to quantify heterogeneity in soft tissue. The images subsequently were compared with histology to identify soft-tissue composition of the plaques.

Methods

Patients with advanced carotid vascular disease requiring endarterectomy were identified and consented to participate. This study did not influence patient selection or treatment. All procedures were approved by the Hines VA Hospital Institutional Review Board. Patients were subjected to routine ultrasound analysis before endarterectomy procedures. After excision, plaques were fixed in formalin and under-went routine pathology. Samples were stored in 10% buffered formalin before imaging. Previous studies have suggested that formalin fixation has minimal effect on tissue properties determined by analyzer-based systems.¹² Ten total plaques were imaged in this study.

The plaque specimens were imaged with a 20-keV x-ray beam by use of a MIR imaging system at the National Synchrotron Light Source in Brookhaven National Laboratory Beamline X 15A as previously described in Zhong et al.¹³ By using a [333] crystal reflection, measurement data were acquired at 21 analyzer crystal positions, ranging from −5 to +5 μrad to create a rocking curve. Image pixel size was 28.1 × 28.1 μm. In addition to performing MIR, synchrotron radiation was used to image the plaques in the absence of the analyzer crystal representing the radiograph used for comparison.

A plaque was imaged further, implementing MIR CT. An x-ray detector (Very High Resolution 1:1, Charge-coupled device, Photonic Science Ltd, Millham, Mountfield, UK) sensor was used to capture x-rays and possessed a detector pixel pitch of 9 μm. The measurement data were acquired at 11 analyzer crystal positions, over 8 μrad using 20-keV x-ray energy. Five hundred tomographic intensity measurements were acquired over 180° angular range for each analyzer-crystal orientation. The acquisition time at each orientation was 1 second.

After imaging, samples were decalcified, dehydrated, and paraffin embedded. Serial sections (5 μm) were cut and alternately stained for hematoxylin and eosin and Masson trichrome through the entire thickness of the plaques. Stained tissue sections were imaged using an Axiovert 200 inverted microscope (Carl Zeiss MicroImaging, Inc, Göttingen, Germany) equipped with a 5 megapixel, 36-bit RGB AxioCam MRc5 color digital camera (Carl Zeiss Microimaging, Inc). The camera and computer-controlled X-Y-Z stage allowed tiling of multiple digital color images to build high-resolution images of entire tissue sections. AxioVision 4.2 image analysis software (Carl Zeiss Microimaging, Inc) was used for acquisition and image processing. Tissue sections were registered to MIR images based on gross morphologic features that are easily identifiable in all MIR images and histologic sections.

For each plaque, 3 MIR images that represent the projected absorption, refraction, and USAXS properties of the plaque were reconstructed as described previously.^14,15 The absorption image is similar to a conventional radiograph; however, it is free of the undesired scattering that usually is present and reduces image quality. The refraction image depicts the effect of small-beam deflections owing to refractive index variations in the object and is determined from the angular shift in the rocking curve distribution when the object is present. The USAXS quantifies angular divergence of the beam caused by the presence of multiple beam refraction from subpixel-sized scatterers and represents broadening of the rocking curve in the presence of the imaged object.

Four images of each plaque were obtained—the radiograph, MIR absorption image, refraction image, and USAXS image. Each image was divided into 3 regions of calcifications, soft tissue, and background. The edge of each plaque was selected manually from the MIR refraction image, in which it was clearly visible (Fig. 1B). Background subtraction was performed on each image.

Figure 1.

MIR Images of plaque (A) absorption, (B) refraction, (C) USAXS, and (D) hematoxylin and eosin histologic section of plaque. Arrows point to interfaces between soft-tissue regions. ^#Calcified regions identified in all 4 images. Circles outline lipid regions.

Six of the radiographs were registered with the MIR images based on the location of calcifications. Calcifications were marked independently in the radiograph for registration with that of the MIR absorption image. It is noted that because all 3 MIR images were acquired simultaneously, no difference in the positioning of the plaque within the images is expected. Only the plaque regions that were available in both the radiograph and the MIR images were used for analysis. Once the corresponding regions were selected in all the images, the appropriate analysis was performed to calculate the metrics for each of the regions in each of the images. Five different regions of interest (150 × 150 pixels) were chosen within the soft-tissue regions and 3 regions were chosen in the background for each of the 6 plaques. Image texture metrics included contrast, energy, uniformity, and entropy.

Statistics

The texture parameters of the MIR absorption, refraction, and USAXS images were compared. The parameters of the MIR absorption images also were compared with the corresponding conventional radiographs. Statistical significance was determined using t tests, with a P value of less than .05 considered significant.

Results

MIR imaging

In all cases, plaques were described as homogeneous and without calcifications based on ultrasound evaluation before harvest. After carotid endarterectomy, the explanted plaques were imaged using MIR. Contrary to the ultrasound analysis, MIR images and histology revealed complex heterogeneous plaque structure with multiple microcalcifications (Fig. 1). The overall shape of the plaque can be seen in the MIR refraction image (Fig. 1B). Also, interfaces between different regions of soft tissue are seen in the refraction image that match the hematoxylin and eosin (Fig. 1D). The microcalcifications can be seen in all 3 MIR images. The USAXS image (Fig. 1C) contains the most contrast in regions of soft tissue, with features generating contrast that are not observed in the absorption image (Fig. 1A). The high-contrast regions distinct from calcified regions appear to correspond to inflammation, necrotic tissue, and lipid-rich regions (Fig. 1D). The amount of heterogeneity (texture) present within the soft tissue of the plaques for each MIR image was quantified by use of statistics from the histograms (uniformity and entropy) and grayco-occurrence matrices (contrast and energy) of regions of interest within the plaque. To show that the observed texture was not caused by noise, these values also were calculated for areas outside of the plaque (background). Four of the texture properties, contrast, uniformity, energy, and entropy are plotted in Fig. 2 for both regions inside and outside the plaque. There was statistical significance for all 4 properties between the soft-tissue regions and background regions. The USAXS image was statistically significantly different from the absorption and refraction images in all 4 categories. The absorption image was significantly different from the refraction image in 3 categories (uniformity, energy, and entropy).

Figure 2.

Image texture was quantified to compare the 3 MIR images. (A) Contrast C = ∑_i,j|i – j|² p(i,j), (B) uniformity $U={\sum }_{i=0}^{L-1}{p}^2\left(z_i\right)$, (C) energyE = ∑_i,jp(i,j)², and (D) entropy $E=-{\sum }_{i=0}^{L-1}p\left(z_i\right){\log }_{2}p\left(z_i\right)$ calculated from 3 MIR images both in soft tissue and background. *P < .05.

Comparison with radiography

MIR absorption images are relatively free of contamination from refraction and scattering. Conventional radiographs also were captured using the same quasimono-chromatic x-ray source at BNL for comparison. These images are better than those available with clinical x-ray systems, but provide a means for comparison with the MIR technique. Synchrotron radiographs only show areas of microcalcification while the MIR absorption image also revealed soft-tissue features (Fig. 3A and B). Texture metrics between the 2 different absorption images were compared. The MIR absorption images were more heterogeneous (P < .05) in all 4 categories within the soft tissue as seen in Fig. 3C–F. These results suggest that the removal of confounding scattering and refraction from x-ray absorption images can provide insight into soft-tissue structure.

Figure 3.

A) Synchrotron radiograph and (B) MIR absorption radiograph of plaque. ^#Calcified regions identified in both images. Circles indicate areas of soft tissue that are visible within the MIR absorption image but not conventional synchrotron absorption image. Image texture of these areas was quantified to compare MIR absorption and conventional x-ray absorption with synchrotron radiation. (C) Contrast, (D) uniformity, (E) energy, and (F) entropy calculated from MIR absorption and synchrotron radiograph both in soft tissue and background. *P < .05.

MIR CT

The MIR planar data suggest that the images revealed more about plaque microstructure than conventional x-ray radiographs and that the refraction and USAXS images produce useful information. However, because the radiographs do not provide details regarding the 3-dimensional (3D) structure of the plaques, CT images also were produced. Overlaying the 3 MIR property images, absorption, refraction, and USAXS, a 3D image of the plaque could be produced (Fig. 4). CT enables imaging of the overall 3D structure of the plaque owing to refraction contrast, and the images allow identification of regions with distinct absorption and USAXS signatures.

Figure 4.

A 3D rendering of MIR CT of carotid artery plaque. Green (light grey) represents refraction, blue (medium grey) represents USAXS, and red (dark grey) represents absorption features. (Color version of the figure is available online).

Comments

This study investigated the use of an analyzer-based x-ray PC imaging method for characterizing human carotid artery plaques. MIR imaging revealed heterogeneous soft-tissue structure not seen in ultrasound and conventional absorption-based x-ray images as well as displaying the calcified regions. Refraction images displayed the edges of the plaque as well as interfaces between different regions of soft tissue. Image texture calculations allowed for the quantification of contrast within different regions of soft tissue. Histology of these regions suggested that soft-tissue regions of necrotic tissue and lipid-rich areas produce specific contrast signatures in the MIR images. Research has shown that these areas may contribute to plaque instability, making them more prone to rupture.^16,17 The ability to identify and calculate the size of these regions noninvasively would be very useful for diagnosis and treatment.

Although this MIR imaging study was performed ex vivo by use of a synchrotron x-ray source, bench-top implementations of analyzer-based PC imaging are under development.¹⁰ This initial example of imaging carotid artery plaque with MIR shows the potential value of the technique. As bench-top and clinical systems are developed, this imaging method shows promise for in vivo imaging of carotid artery plaques.

There is a significant need for imaging methods that are able to identify asymptomatic plaques that are at high risk for stroke. Our results indicate that MIR imaging has the potential to identify soft-tissue structures in carotid plaques. Future research needs to focus on computed tomography implementation of MIR to quantify texture and match PC CT images with corresponding histology in both symptomatic and asymptomatic plaques. This research would determine the specificity and sensitivity of MIR for identification of distinct areas of calcifications, lipid-rich regions, necrotic tissue, inflammation, and fibrous regions. This information subsequently could be applied toward the evaluation of x-ray PC CT imaging for identifying asymptomatic high-risk plaques by identification of 3-dimensional tissue composition and organization. This preliminary research shows the MIR imaging technique can identify carotid plaque microstructure and should be investigated further as a viable imaging technique to identify high-risk plaques.