Adoption of a new automated optical coherence tomography software to obtain a lipid plaque spread-out plot

Background

Near infrared spectroscopy – intravascular ultrasound (NIRS-IVUS) imaging can provide a fully automated estimation of lipid burden, providing a two-dimensional spread-out plot, the Lipid Core Burden Index (LCBI), which has been associated with higher incidence of cardiac events. Optical coherence tomography (OCT) can identify lipid component with high accuracy and it is therefore potentially capable of measuring its longitudinal extension in a dedicated two-dimensional LCBI spread-out plot.

Purpose

The present study has been designed to validate a novel automated approach to assess OCT images, able of providing a dedicated LCBI spread-out plot plus other features of plaque vulnerability.

Methods

We compared the results obtained with a novel automated OCT alghorithm, developed utilising a convolutional neural network, with those obtained with conventional (manual) OCT and with NIRS-IVUS in a consecutive series of 40 patients with coronary artery disease. We tested and validated our new OCT algorithm to calculate the lipid core longitudinal extension in a dedicated two-dimensional LCBI spread-out plot. In each coronary plaque, the following measurements were obtained with NIRS-IVUS: 1) minimum lumen area (MLA), 2) vessel area at MLA site, 3) plaque burden (%) at MLA site, 4) NIRS-defined lipid pool arch and 5) maximum LCBI measurement within a 4 mm length. The following OCT features were obtained: 1) the MLA cross section, 2) the minimum fibrous cap thickness (FCT) in presence of lipid components and measured as the average of three measurements obtained in the same cross-section and 3) maximum LCBI within a 4 mm length.

Results

Three lesions groups were identified according to the studied lesions: 1) culprit lesions in patients with acute coronary syndrome (ACS, n=16), 2) non-culprit lesions in patients with ACS (n=12) and 3) lesions in patients with stable angina (n=12). OCT conventional assessment showed for the culprit ACS plaques a trend for a larger lipid arc and a significant thinner FCT (p=0.028). Consistently, NIRS-IVUS showed for culprit ACS plaques a more complex anatomy. A strong trend for increased maximum LPBI in 4mm segments was found in the culprit ACS group, regardless of the adopted imaging modality, either NIRS-IVUS or automated OCT (p=0.184 and p=0.066, respectively, figure 1). A fair correlation was obtained for the maximum 4 mm LCBI measured by NIRS-IVUS and automated OCT (r=0.75). The sensitivity and specificity of automated OCT to detect significant LCBI, applying a validated 400 cut off were 90.5 and 84.2 respectively.

Conclusions

We developed an automated approach, comparable to NIRS, to assess OCT images that can provide a dedicated lipid plaque spread-out plot to address plaque vulnerability. The automated OCT software can promote and improve OCT clinical applications for the identification of patients at risk of hard events.

Funding Acknowledgement

Type of funding sources: Foundation. Main funding source(s): CLI - Centro Lotta all'Infarto Spread-out plot by IVUS-NIRS and OCT

Use of enhanced stent visualisation compared to angiography alone to guide percutaneous coronary intervention

OBJECTIVE

We aimed to assess the use of enhanced stent visualisation (ESV) on outcomes, after PCI with overlapping stents, specifically using CLEARstent technology.

BACKGROUND

Stent underexpansion and overlap are both significant risk factors for restenosis and stent thrombosis. Enhanced stent visualisation (e.g. CLEARstent) systems could provide important data to reduce under-expansion and stent overlap.

METHODS

This was a cohort study based on this institution's percutaneous coronary intervention (PCI) registry. A total of 2614 patients who had PCI for stable angina or acute coronary syndromes (ACS, excluding cardiogenic shock) with overlapping 2nd generation drug eluting stents (DES) in the same vessel between May 2015 and January 2018 were included in the analysis. Patients were divided into ESV (n = 1354) and no ESV guided intervention (n = 1260). The primary end-point was major adverse cardiovascular events (MACE: target vessel revascularisation, target vessel myocardial infarction and all-cause mortality) recorded at a median follow up of 2.4 years.

RESULTS

Groups were comparable for patient characteristics (age, diabetes mellitus, ACS presentation). A significant difference in MACE was observed between patients who underwent ESV-guided PCI (9.5%) compared with patients who underwent Standard PCI (14.4%, p = .018). This difference was mainly driven by reduced rates of target vessel revascularisation and recurrent myocardial infarction. Overall this difference persisted after multivariate Cox analysis (HR 0.86, 95% CI: 0.73-0.98) and propensity matching (HR = 0.88, 95% CI: 0.69-0.99).

CONCLUSION

We suggest that routine clinical use of ESV technology during PCI can be useful, and is associated with better medium-term angiographic and clinical outcomes. Further study is required to build on this promising signal.

P2720Diagnostic accuracy of Quantitative Flow Ratio (QFR) and Vessel Fractional Flow Reserve (vFFR) compared to Fractional Flow Reserve (FFR) based on 7.5 frames/second coronary angiography

Background

Fractional flow reserve (FFR) is the gold standard for the physiological assessment of intermediate coronary artery lesions. Recently, several novel methods for computation of FFR based on 3-dimensional quantitative coronary angiography have been developed. These techniques allow analyses to be performed retrospectively and do not require induction of hyperaemia. The development and validation of these techniques are based on good quality coronary angiography with high frames per second (15 fps) acquisition. The diagnostic accuracy of Quantitative Flow Ratio (QFR) and Vessel Fractional Flow Reserve (vFFR) in real world “radiation-save mode” coronary angiography has not been studied.

Purpose

To validate the accuracy of QFR and vFFR compared to FFR based on a series of coronary angiography acquired at 7.5 fps.

Methods

We retrospectively analyzed 134 vessels (102 patients) with intermediate coronary artery stenosis (30–90%) in whom an FFR measurement had been performed. All the coronary angiography were acquired at 7.5 fps. 33 vessels (20 patients) were excluded from the study due to unsuitable coronary anatomy, invalid FFR measurements, poor image quality and lack of 2 projections ≥25° apart. A total of 101 vessels (82 patients) were included in the final analysis. Contrast-QFR (cQFR), fixed-QFR (fQFR) and vFFR analysis were performed in these vessels by two independent trained experts blinded to the FFR readings. FFR measurements at hyperaemic steady state was taken as the gold standard reference.

Results

Good intra- and inter-observer reliability was noted for fQFR, cQFR and vFFR analysis (intra-observer mean difference for fQFR: 0.016±0.060, p=0.066; cQFR: 0.009±0.053, p=0.230; vFFR: 0.008±0.040, p=0.175; inter-observer mean difference for fQFR: 0.001±0.036, p=0.847; cQFR: −0.001±0.049; p=0.910, vFFR: −0.005±0.037, p=0.393). fQFR and cQFR showed good correlation with FFR (r=0.694, p<0.001 and r=0.674, p<0.001, respectively) while vFFR showed moderate correlation with FFR (r=0.388, p<0.001). Similarly, fQFR and cQFR showed good accuracy for the detection of functionally significant coronary stenosis (fQFR AUC 0.882 (95% CI 0.803–0.938) and cQFR AUC 0.886 (95% CI 0.807–0.940)) while vFFR showed moderate accuracy with AUC 0.719 (95% CI 0.621–0.804). For identifying functionally significant stenosis (FFR ≤0.80), the overall diagnostic accuracy were 81.2%, 85.2%, 75.3% for fQFR, cQFR and vFFR, repectively. The sensitivity and specificity were 72.7%, 89.9% (fQFR); 83.5%, 31.8% (cQFR) and 68.2%, 87.3% (vFFR).

Conclusion

Functional assessment of intermediate coronary stenosis based on 7.5 fps angiography-derived computational modelling is feasible. Our study shows that fQFR and cQFR have a better diagnostic accuracy for detecting functionally significant coronary stenosis compared to vFFR. At the lower radiation-save mode 7.5 fps angiography, cQFR does not appear to provide additional diagnostic accuracy compared to fQFR.

3D reconstruction of coronary arteries using frequency domain optical coherence tomography images and biplane angiography

The aim of this study is to describe a new method for three-dimensional (3D) reconstruction of coronary arteries using Frequency Domain Optical Coherence Tomography (FD-OCT) images. The rationale is to fuse the information about the curvature of the artery, derived from biplane angiographies, with the information regarding the lumen wall, which is produced from the FD-OCT examination. The method is based on a three step approach. In the first step the lumen borders in FD-OCT images are detected. In the second step a 3D curve is produced using the center line of the vessel from the two biplane projections. Finally in the third step the detected lumen borders are placed perpendicularly onto the path based on the centroid of each lumen border. The result is a 3D reconstructed artery produced by all the lumen borders of the FD-OCT pullback representing the 3D arterial geometry of the vessel.

In vivovalidation of a novel semi-automated method for border detection in intravascular ultrasound images

The aim of this work was to evaluate a new semi-automated intravascular ultrasound (IVUS) border detection method. The method was used to identify the lumen and the external elastic membrane or the borders of stents in 80 IVUS images, randomly selected from 10 consecutive human coronary arteries. These semi-automated results were compared with observations of two experts. Several indices in each case were obtained in order fully to evaluate the method. The time required for identification of the borders was also recorded. The interobserver variability of the method ranged from 1.21% to 5.61%, the correlation coefficient from 0.98 to 0.99, the slope was close to unity (0.94-1.03), the y intercept close to zero and the Williams index value was close to unity (range 0.67-0.91). The time (mean+/-SD) required for the method to identify the borders of the different vessel layers for the whole IVUS sequence was 5.2+/-0.2 min. The results demonstrate that the method is reliable and capable of identifying rapidly and accurately the different vessel layers depicted in IVUS images.