1. Introduction
Coronary angiography is the primary diagnostic imaging modality for the evaluation and classification of coronary artery lesions as well as for guiding percutaneous coronary interventions (PCIs). Percutaneous interventions are the most performed coronary revascularization procedure, improving the quality of life of patients along with their clinical outcomes [
1]. Despite major advances in coronary stent technology, acute and late PCI-related complications still occur [
2,
3,
4]. Successful PCI results relate directly to proper stent placement and deployment. Stent underexpansion was shown to be a major predictor of stent restenosis and thrombosis by quantitative coronary angiography (QCA) [
5]. Moreover, insufficient stent expansion and malapposition found by intracoronary imaging were shown as major predictors of stent thrombosis in several studies [
6,
7,
8,
9,
10].
Although optimizing stent implantation under intravascular imaging guidance is widely supported by the current literature [
11,
12,
13,
14,
15,
16,
17], its routine clinical use remains limited due to the added time and cost to the procedure along with the image interpretation difficulties. Despite conventional angiography often falling behind in the detection of stent underexpansion and presenting a suboptimal accuracy assessing stent position, it is still carried out during routine clinical practice especially with newer generation scaffolds that are implanted at a higher pressure followed by a post-dilatation step and rely on the radiopaque nature of the material used for visualization.
Thicker stent struts were associated with higher in-stent restenosis rates in the ISAR STEREO trials [
18,
19]. On the other hand, thinner strut scaffolds used in new generation stents have been advocated to significantly reduce the risk of myocardial infarction at the expense of being more radiolucent on fluoroscopy [
3,
20,
21,
22,
23]. Moreover, the trend towards the use of lower X-ray power during angiographic procedures presents another challenge for stent visualization which is further altered due to motion during the angiography sequence secondary to X-ray scattering.
More recently, several enhanced stent imaging (ESI) methods have been developed. These angiography-based software improve stent visualization and provide quantitative as well as qualitative data post-stent deployment but remain dependent on the X-ray angiographic system of each vendor [
24,
25]. The StentBoost
® system (SB) (Philips Healthcare, Andover, MA, USA) is a motion-corrected X-ray stent visualization software that allows better assessment of stent expansion without using contrast [
25]. It was designed as an add-on to conventional X-ray angiographic system and was found to be superior to conventional angiography in detecting stent underexpansion. The algorithm relies on the motion-compensated noise reduction by using landmarks (balloon markers) on 45 registered frames acquired over 3–4 s [
26]. These images are transferred automatically to a workstation and corrected by averaging the images from each cine frame in relation to the two balloon markers. The software enhances stent visibility, fading out anatomical structures and background noise [
25]. SB was found to have good correlation with IVUS regarding stent diameter and was found to be superior to quantitative coronary angiography (QCA) [
14,
17,
24,
27,
28,
29,
30,
31].
Pie Medical Imaging (Maastricht, The Netherlands) introduced the CAAS StentEnhancer® (SE), a method similar to SB with the main advantage of being completely independent of the X-ray angiographic system of the vendor and hence, runs on a side station. SE uses a maximum of 40 frames from a Digital Imaging and Communications in Medicine (DICOM) file. Its algorithm automatically detects the markers of the stent balloon or of the balloon used for post-dilation in order to compute a single image in which the visibility of a deployed stent is improved. Following background subtraction, all frames are transformed into a common reference frame. The resulting images are combined into a single image after weighted averaging. A sharpening filter is then applied. This filter works by first extracting the high-frequency components from the image. These high-frequency components are then added, using a predefined amount, to the original image. High-frequency components are extracted by first creating a blurred version of the image through performing a convolution of a Gaussian filter at a predefined scale with the original image. Subtracting the blurred version from the original yields the high-frequency components. An optimally contrasted enhanced stent image is then generated to improve the visibility using a linear scaling within a predefined width around the peak pixel value which is established from a histogram analysis. Furthermore, the SE system allows for a manual contrast adjustment of the generated images as well as a quantitative assessment of the deployed stent through manual measurements of different diameters along the stent length.
Quantitative coronary angiography (QCA) is a tool to measure coronary arteries filled with contrast based on the use of a dedicated software allowing automated measurements (that can be manually corrected) of vessel diameter, percent stenosis, and minimal lumen of stent diameters [
32]. After image acquisition, a digital quantification on a selected frame can be easily performed with or without magnification.
The aims of this study was to (1) qualitatively compare image results from the SE system to the currently available SB system and (2) report the comparisons between measured diameters of deployed stents by the SE system and the expected chart diameters upon deployment and after post-dilation as well as final QCA measurements.
2. Materials and Methods
2.1. Study Design
Between January 2016 and January 2018, patients in whom an ESI acquisition was performed after the implantation of a stent at the Centre Hospitalier Universitaire et Psychiatrique de Mons-Borinage (CHUPMB), Belgium, were retrospectively reviewed. The acquired ESI images were transferred to the SB and SE workstations (CAAS workstation software v.8.4) and reconstructed. The patients’ baseline demographic and procedural characteristics were collected. The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of CHUPMB and Erasme-ULB (Université Libre de Bruxelles) (protocol code P2017/462 on 16 October 2017) who waived the requirement for written consent. Two independent blinded and experienced interventional cardiologists compared and graded the stent images obtained by each technique. The images from the same sequence (SB and SE) were blindly compared side to side (on the left side, it was either a SB or SE image and on the right side the other one). The images were graded on a scale from 0 to 2 (0 = undetectable; 1 = seen unclearly; and 2 = clearly seen) for different characteristics: (1) visualization of the proximal edge of the stent; (2) visualization of the distal edge of the stent; (3) clear visualization of the struts of the stent; and (4) the presence of underexpansion and (5) calcifications. One month later, 50 sequences were randomly selected and re-analyzed a second time by one of the two observers for the intra-observer variation analysis.
A subset of images was processed using a custom-designed Matlab software (version R2017a, The MathWorks, Natick, MA, USA) that computed the signal-to-noise ratio (SNR) of SB and SE images defined as the ratio of the average signal value μ
sig to the standard deviation σ
bg of the background. As shown in
Figure 1, a reference noise square of 100 by 100 pixels was manually placed in a region without interventional material (wire, previous stent, etc.) and without a bone structure such as a rib. Another rectangle was then traced around the stent, as close as possible to the struts. The same two regions were used in the SB and SE images for comparison. The standard deviation σ
bg of the background pixel values was calculated in the square region of interest (ROI) of noise whereas the average signal value μ
sig was calculated as the average of the values of the pixels in the ROI traced around the enhanced stent.
Furthermore, between January 2021 and July 2022, patients with mildly to moderately calcified de novo coronary lesions in 4 Belgian centers who were treated by stent implantation and ESI acquisition in 2 orthogonal views were prospectively included. This protocol with EudraCT code B7072020000065 was approved by the Ethics Committee Hospitalo-Facultaire Universitaire de Liège under reference 2020/87 on 13/11/2020, as well as by each local institution review board. The study was also conducted according to the guidelines of the Declaration of Helsinki and informed written consent were obtained. The patients’ baseline demographic and procedural characteristics were collected. The ESI images were transferred to the SE and QCA workstations (CAAS software v.8.4) and reconstructed. Of note, one center used a Siemens X-ray system with the Clearstent ESI system, the others used a Philips system with StentBoost. SE and QCA could be measured on the DICOM files from these two manufacturers. The final QCA analysis was conducted and included maximal and minimal stent diameters as well as percent stenosis. A quantitative SE analysis of the 2 orthogonal views acquired including proximal and distal stent edge diameters as well as minimal stent diameter was conducted. Mean stent diameter as well as percent stenosis were calculated in both views and compared to the expected stent chart diameter according to the pressure of deployment of the stent and after post-dilatation when available as well as to the QCA measurements. A Bland–Altman analysis was performed to compare the SE and QCA diameters.
2.2. Statistical Analysis
Categorical variables are reported as absolute values and percentages. Continuous variables are presented as means and standard deviations.
The Wilcoxon test was used to compare the two software and the two observers. After comparisons of the two methods, Kappa coefficients were calculated for repeatability and agreement between the reviewers.
The Kendall test was used to compare the two software for the presence of calcifications and stent underexpansion. Two McNemar tests were used for the evaluation of calcifications on underexpansion and post-dilatation efficacy. The SNRs of the SB and SE images were compared using a paired t-test. A p value < 0.05 was considered statistically significant. All statistical analyses was performed using SPSS software v.23 (IBM, New York, NY, USA).
4. Discussion
The current study qualitatively compared the inter- and intra-observer results of different image criteria of a novel ESI software (SE) to the market-available one (SB) as well as a quantitative analysis of SE.
The results of the two ESI algorithms were compared as per each observer and finally SNRs for the two methods were calculated and compared to the SNR calculated from the angiographic image. ESI methods have been demonstrated to enhance contrast on fluoroscopic images, allowing better visualization of stent struts. This study demonstrated that SE is not inferior to SB for the criteria evaluated but a clear inter-observer variability calls for more quantitative methods. Despite this difference, both observers had a preference towards SE images to study parameters. SE can be easily integrated into procedures, independent of the X-ray angiography machine vendor, and was found in our study to provide good stent expansion assessment as well as a better stent strut visualization. The SNR of SE images was found to be superior compared to SB. We cannot provide a definite answer why this was the case, since we are unaware of the exact methods of StentBoost. However, based on the available papers, there are methodological differences. For instance, SB does not seem to perform background subtraction as can be seen in
Figure 1.
Since newer generation scaffolds tend to use thinner struts or bioresobable materials in order to reduce the risk of stent thrombosis in addition to a trend towards the use of lower X-ray power during angiographic procedures, proper stent visualization is becoming challenging [
3,
20,
21,
22,
23]. The use of ESI becomes pivotal for the assessment of proper stent expansion, a major risk factor for stent thrombosis [
6,
7,
8,
9,
10].
QCA remains an important, readily available, and easy-to-use tool during PCIs allowing for a more practical and standardized angiography-based approach. QCA is particularly useful for the evaluation of the minimal lumen diameter, the reference vessel diameter, the diameter stenosis percentage, the lesion length, the acute gain, and late loss [
14]. Our data failed to show any correlation between the expected stent diameter and the QCA-derived one. This could be explained by the foreshortening drawback of QCA as well as the two-dimensional evaluation by QCA of a three-dimensional vessel.
Our data demonstrated the feasibility of an accurate, quantitative, contrast-free assessment of stent expansion by SE. Post-dilatation remains an important step towards stent optimization. Our results are in-line with the current published literature in regard to the achieved stent diameter at a given implantation pressure being at least 10% lower than the given expected chart diameter [
33,
34]. These results could be attributed to the fact that the figures provided on the compliance charts are derived from bench tests performed in water at 37 °C while QCA and SE are measured on stents deployed in fibrotic and calcified lesions. The diameters measured by SE remain 2-dimensional measurements of a 3-dimensional structure. A second measurement using an orthogonal view would therefore overcome this limitation. The measured mean stent diameters by SE did not differ when using the two orthogonal views indicating precise measurements between the two views. However, a difference was noted for the measurements at the site of maximal stent underexpansion after post-dilation. This could be attributed to the eccentricity of lesions as well as to the visual assessment of the minimal diameter compared to adjacent ones.
Despite an era where modern flat-detector technology allows excellent angiographic images, coronary stent visualization has become a challenge especially with the on-going reduction in stent strut thickness. Stents are often suboptimally visualized on plain angiography hence limiting optimal PCI outcomes. High temporal resolution is needed to qualify a moving structure.
Adequate stent expansion has important short- and long-term effects after PCIs in clinical practice. It is crucial, yet challenging, to detect suboptimal stent deployment on qualitative and quantitative angiography since it is associated with an increased rate of in-stent restenosis and stent thrombosis [
3,
20]. Current stent delivery systems are still suboptimal for stent expansion, requiring, in most cases, a post-dilation using a larger, higher pressure, non-compliant balloon to improve the in-lumen area. This is particularly true when increased calcifications are found [
24,
35] as we demonstrated in our 157 lesions.
Intracoronary imaging, including intravascular ultrasound and optical coherence tomography, remains more sensitive than angiography and QCA in determining stent under expansion as illustrated in
Figure 4; their use was found to improve stent expansion results and long term outcome [
6,
20,
36]. However, these techniques are limited by cost, time, and technical expertise, calling for a simpler, ready-to-use visualization method. ESI was found very useful in identifying stents under expansion, thereby improving PCI outcomes [
37]. Image processing algorithm softwares based on X-ray angiography images offer better stent visualization compared to angiography alone as validated by several previous studies [
24,
38]. ESI also shows no risk of complications and adds little additional time or radiation to the procedure [
25,
37]. Furthermore, ESI allows accurate measurements of the dimension of stents [
29]. It was found useful in obese patients, long lesions, in-stent restenosis, and bifurcating lesions. Moreover, ESI was found to be superior to QCA and angiography and was highly correlated with IVUS [
17,
28,
28,
30,
38].
We performed a comparison between the final diameters measured by QCA and by SE. We had 14 paired data available with no post-dilatation and 75 after final post-dilatation in the other patients. As shown in
Figure 5, on average, the mean difference was −0.37 ± 0.42 mm and this bias for smaller QCA diameters was significant, with the 95% confidence interval not encompassing the 0. Using IVUS as the reference, Goto et al. also demonstrated that QCA underestimates MLDs in small vessels (<3.8 mm) and overestimates MLDs in vessels larger than 3.8 mm [
39]. A direct head-to-head comparison of IVUS and SE might confirm a better agreement between IVUS and SE than with QCA. The wide agreement window between QCA and SE of ± 0.84 mm reflects the differences in the two methodologies, with only manual measurements being currently possible with the SE images, while automated contour calculation and minimal and reference diameters are available with QCA. However, when there were overlap of side branches or other vessels, no QCA could be reliably measured in 3 out of the 93 cases. Without contrast, hence without any overlap, SE could always be measured. Of note, no reliable reconstruction can be calculated on very long stents and/or with long balloons when the markers are more than 30 mm apart. Several ESI softwares are currently available on the market but each one can only be used on the specific vendor’s angiographic system. The StentEnhancer software computes enhanced fluoroscopy images using the balloon markers as references, delivering an easily integrated, high-quality image independent of the angiography instrument vendor. Although an increase in radiation was reported during ESI acquisition, no significant impact on the patient radiation dose was found [
40].