CARDIAN: a novel computational approach for real-time end-diastolic frame detection in intravascular ultrasound using bidirectional attention networks

Introduction

Changes in coronary artery luminal dimensions during the cardiac cycle can impact the accurate quantification of volumetric analyses in intravascular ultrasound (IVUS) image studies. Accurate ED-frame detection is pivotal for guiding interventional decisions, optimizing therapeutic interventions, and ensuring standardized volumetric analysis in research studies. Images acquired at different phases of the cardiac cycle may also lead to inaccurate quantification of atheroma volume due to the longitudinal motion of the catheter in relation to the vessel. As IVUS images are acquired throughout the cardiac cycle, end-diastolic frames are typically identified retrospectively by human analysts to minimize motion artefacts and enable more accurate and reproducible volumetric analysis.

Methods

In this paper, a novel neural network-based approach for accurate end-diastolic frame detection in IVUS sequences is proposed, trained using electrocardiogram (ECG) signals acquired synchronously during IVUS acquisition. The framework integrates dedicated motion encoders and a bidirectional attention recurrent network (BARNet) with a temporal difference encoder to extract frame-by-frame motion features corresponding to the phases of the cardiac cycle. In addition, a spatiotemporal rotation encoder is included to capture the IVUS catheter's rotational movement with respect to the coronary artery.

Results

With a prediction tolerance range of 66.7 ms, the proposed approach was able to find 71.9%, 67.8%, and 69.9% of end-diastolic frames in the left anterior descending, left circumflex and right coronary arteries, respectively, when tested against ECG estimations. When the result was compared with two expert analysts' estimation, the approach achieved a superior performance.

Discussion

These findings indicate that the developed methodology is accurate and fully reproducible and therefore it should be preferred over experts for end-diastolic frame detection in IVUS sequences.

Morphological Changes of Anomalous Coronary Arteries From the Aorta During the Cardiac Cycle Assessed by IVUS in Resting Conditions

BACKGROUND

Anomalous aortic origin of coronary artery (AAOCA) with intramural segment is associated with risk of sudden cardiac death, probably related to a compressive mechanism exerted by the aorta. However, the intramural compression occurrence and magnitude during the cardiac cycle remain unknown. We hypothesized that (1) in end diastole, the intramural segment is narrower, more elliptic, and has greater resistance than extramural segment; (2) the intramural segment experiences a further compression in systole; and (3) morphometry and its systolic changes vary within different lumen cross-sections of the intramural segment.

METHODS

Phasic changes of lumen cross-sectional coronary area, roundness (minimum/maximum lumen diameter), and hemodynamic resistance (Poiseuille law for noncircular sections) were derived from intravascular ultrasound pullbacks at rest for the ostial, distal intramural, and extramural segments. Data were obtained for 35 AAOCA (n=23 with intramural tract) after retrospective image-based gating and manual lumen segmentation. Differences between systolic and end-diastolic phases in each section, between sections of the same coronary, and between AAOCA with and without intramural tract were assessed by nonparametric statistical tests.

RESULTS

In end diastole, both the ostial and distal intramural sections were more elliptical (P<0.001) than the reference extramural section and the correspondent sections in AAOCA without intramural segment. In systole, AAOCA with intramural segment showed a flattening at the ostium (-6.76% [10.82%]; P=0.024) and a flattening (-5.36% [16.56%]; P=0.011), a narrowing (-4.62% [11.38%]; P=0.020), and a resistance increase (15.61% [30.07%]; P=0.012) at the distal intramural section. No-intramural sections did not show morphological changes during the entire cardiac cycle.

CONCLUSIONS

AAOCA with intramural segment has pathological segment-specific dynamic compression mainly in the systole under resting conditions. Studying AAOCA behavior with intravascular ultrasound during the cardiac cycle may help to evaluate and quantify the severity of the narrowing.

A Deep Learning Method for Motion Artifact Correction in Intravascular Photoacoustic Image Sequence

In vivo application of intravascular photoacoustic (IVPA) imaging for coronary arteries is hampered by motion artifacts associated with the cardiac cycle. Gating is a common strategy to mitigate motion artifacts. However, a large amount of diagnostically valuable information might be lost due to one frame per cycle. In this work, we present a deep learning-based method for directly correcting motion artifacts in non-gated IVPA pullback sequences. The raw signal frames are classified into dynamic and static frames by clustering. Then, a neural network named Motion Artifact Correction (MAC)-Net is designed to correct motion in dynamic frames. Given the lack of the ground truth information on the underlying dynamics of coronary arteries, we trained and tested the network using a computer-generated dataset. Based on the results, it has been observed that the trained network can directly correct motion in successive frames while preserving the original structures without discarding any frames. The improvement in the visual effect of the longitudinal view has been demonstrated based on quantitative evaluation of the inter-frame dissimilarity. The comparison results validated the motion-suppression ability of our method comparable to gating and image registration-based non-learning methods, while maintaining the integrity of the pullbacks without image preprocessing. Experimental results from in vivo intravascular ultrasound and optical coherence tomography pullbacks validated the feasibility of our method in the in vivo intracoronary imaging scenario.

Quantifying Patient-Specific in vivo Coronary Plaque Material Properties for Accurate Stress/Strain Calculations: An IVUS-Based Multi-Patient Study

Introduction: Mechanical forces are closely associated with plaque progression and rupture. Precise quantifications of biomechanical conditions using in vivo image-based computational models depend heavily on the accurate estimation of patient-specific plaque mechanical properties. Currently, mechanical experiments are commonly performed on ex vivo cardiovascular tissues to determine plaque material properties. Patient-specific in vivo coronary material properties are scarce in the existing literature.

Methods:In vivo Cine intravascular ultrasound and virtual histology intravascular ultrasound (IVUS) slices were acquired at 20 plaque sites from 13 patients. A three-dimensional thin-slice structure-only model was constructed for each slice to obtain patient-specific in vivo material parameter values following an iterative scheme. Effective Young's modulus (YM) was calculated to indicate plaque stiffness for easy comparison purposes. IVUS-based 3D thin-slice models using in vivo and ex vivo material properties were constructed to investigate their impacts on plaque wall stress/strain (PWS/PWSn) calculations.

Results: The average YM values in the axial and circumferential directions for the 20 plaque slices were 599.5 and 1,042.8 kPa, respectively, 36.1% lower than those from published ex vivo data. The YM values in the circumferential direction of the softest and stiffest plaques were 103.4 and 2,317.3 kPa, respectively. The relative difference of mean PWSn on lumen using the in vivo and ex vivo material properties could be as high as 431%, while the relative difference of mean PWS was much lower, about 3.07% on average.

Conclusion: There is a large inter-patient and intra-patient variability in the in vivo plaque material properties. In vivo material properties have a great impact on plaque stress/strain calculations. In vivo plaque material properties have a greater impact on strain calculations. Large-scale-patient studies are needed to further verify our findings.

IVUS Longitudinal and Axial Registration for Atherosclerosis Progression Evaluation

Intravascular ultrasound (IVUS) imaging offers accurate cross-sectional vessel information. To this end, registering temporal IVUS pullbacks acquired at two time points can assist the clinicians to accurately assess pathophysiological changes in the vessels, disease progression and the effect of the treatment intervention. In this paper, we present a novel two-stage registration framework for aligning pairs of longitudinal and axial IVUS pullbacks. Initially, we use a Dynamic Time Warping (DTW)-based algorithm to align the pullbacks in a temporal fashion. Subsequently, an intensity-based registration method, that utilizes a variant of the Harmony Search optimizer to register each matched pair of the pullbacks by maximizing their Mutual Information, is applied. The presented method is fully automated and only required two single global image-based measurements, unlike other methods that require extraction of morphology-based features. The data used includes 42 synthetically generated pullback pairs, achieving an alignment error of 0.1853 frames per pullback, a rotation error 0.93° and a translation error of 0.0161 mm. In addition, it was also tested on 11 baseline and follow-up, and 10 baseline and post-stent deployment real IVUS pullback pairs from two clinical centres, achieving an alignment error of 4.3±3.9 for the longitudinal registration, and a distance and a rotational error of 0.56±0.323 mm and 12.4°±10.5°, respectively, for the axial registration. Although the performance of the proposed method does not match that of the state-of-the-art, our method relies on computationally lighter steps for its computations, which is crucial in real-time applications. On the other hand, the proposed method performs even or better that the state-of-the-art when considering the axial registration. The results indicate that the proposed method can support clinical decision making and diagnosis based on sequential imaging examinations.

End-diastolic segmentation of intravascular ultrasound images enables more reproducible volumetric analysis of atheroma burden

BACKGROUND

Volumetric intravascular ultrasound (IVUS) analysis is currently performed at a fixed frame interval, neglecting the cyclic changes in vessel dimensions occurring during the cardiac cycle that can affect the reproducibility of the results. Analysis of end-diastolic (ED) IVUS frames has been proposed to overcome this limitation. However, at present, there is lack of data to support its superiority over conventional IVUS.

OBJECTIVES

The present study aims to compare the reproducibility of IVUS volumetric analysis performed at a fixed frame interval and at the ED frames, identified retrospectively using a novel deep-learning methodology.

METHODS

IVUS data acquired from 97 vessels were included in the present study; each vessel was segmented at 1 mm interval (conventional approach) and at ED frame twice by an expert analyst. Reproducibility was tested for the following metrics; normalized lumen, vessel and total atheroma volume (TAV), and percent atheroma volume (PAV).

RESULTS

The mean length of the analyzed segments was 50.0 ± 24.1 mm. ED analysis was more reproducible than the conventional analysis for the normalized lumen (mean difference: 0.76 ± 4.03 mm³ vs. 1.72 ± 11.37 mm³ ; p for the variance of differences ratio < 0.001), vessel (0.30 ± 1.79 mm³ vs. -0.47 ± 10.26 mm³ ; p < 0.001), TAV (-0.46 ± 4.03 mm³ vs. -2.19 ± 14.39 mm³ ; p < 0.001) and PAV (-0.12 ± 0.59% vs. -0.34 ± 1.34%; p < 0.001). Results were similar when the analysis focused on the 10 mm most diseased segment. The superiority of the ED approach was due to a more reproducible detection of the segment of interest and to the fact that it was not susceptible to the longitudinal motion of the IVUS probe and the cyclic changes in vessel dimensions during the cardiac cycle.

CONCLUSIONS

ED IVUS segmentation enables more reproducible volumetric analysis and quantification of TAV and PAV that are established end points in longitudinal studies assessing the efficacy of novel pharmacotherapies. Therefore, it should be preferred over conventional IVUS analysis as its higher reproducibility is expected to have an impact on the sample size calculation for the primary end point.

Coronary arterial geometry: A comprehensive comparison of two imaging modalities

The characterization of vascular geometry is a fundamental step towards the correct interpretation of coronary artery disease. In this work, we report a comprehensive comparison of the geometry featured by coronary vessels as obtained from coronary computed tomography angiography (CCTA) and the combination of intravascular ultrasound (IVUS) with bi-plane angiography (AX) modalities. We analyzed 34 vessels from 28 patients with coronary disease, which were deferred to CCTA and IVUS procedures. We discuss agreement and discrepancies between several geometric indexes extracted from vascular geometries. Such an analysis allows us to understand to which extent the coronary vascular geometry can be reliable in the interpretation of geometric risk factors, and as a surrogate to characterize coronary artery disease.

A deep learning methodology for the automated detection of end-diastolic frames in intravascular ultrasound images

Coronary luminal dimensions change during the cardiac cycle. However, contemporary volumetric intravascular ultrasound (IVUS) analysis is performed in non-gated images as existing methods to acquire gated or to retrospectively gate IVUS images have failed to dominate in research. We developed a novel deep learning (DL)-methodology for end-diastolic frame detection in IVUS and compared its efficacy against expert analysts and a previously established methodology using electrocardiographic (ECG)-estimations as reference standard. Near-infrared spectroscopy-IVUS (NIRS-IVUS) data were prospectively acquired from 20 coronary arteries and co-registered with the concurrent ECG-signal to identify end-diastolic frames. A DL-methodology which takes advantage of changes in intensity of corresponding pixels in consecutive NIRS-IVUS frames and consists of a network model designed in a bidirectional gated-recurrent-unit (Bi-GRU) structure was trained to detect end-diastolic frames. The efficacy of the DL-methodology in identifying end-diastolic frames was compared with two expert analysts and a conventional image-based (CIB)-methodology that relies on detecting vessel movement to estimate phases of the cardiac cycle. A window of ± 100 ms from the ECG estimations was used to define accurate end-diastolic frames detection. The ECG-signal identified 3,167 end-diastolic frames. The mean difference between DL and ECG estimations was 3 ± 112 ms while the mean differences between the 1st-analyst and ECG, 2nd-analyst and ECG and CIB-methodology and ECG were 86 ± 192 ms, 78 ± 183 ms and 59 ± 207 ms, respectively. The DL-methodology was able to accurately detect 80.4%, while the two analysts and the CIB-methodology detected 39.0%, 43.4% and 42.8% of end-diastolic frames, respectively (P < 0.05). The DL-methodology can identify NIRS-IVUS end-diastolic frames accurately and should be preferred over expert analysts and CIB-methodologies, which have limited efficacy.

A pragmatic approach to understand peripheral artery lumen surface stiffness due to plaque heterogeneity

The goal of this study was to develop a pragmatic approach to build patient-specific models of the peripheral artery that are aware of plaque inhomogeneity. Patient-specific models using element-specific material definition (to understand the role of plaque composition) and homogeneous material definition (to understand the role of artery diameter and thickness) were automatically built from intravascular ultrasound images of three artery segments classified with low, average, and high calcification. The element-specific material models had average surface stiffness values of 0.0735, 0.0826, and 0.0973 MPa/mm, whereas the homogeneous material models had average surface stiffness values of 0.1392, 0.1276, and 0.1922 MPa/mm for low, average, and high calcification, respectively. Localization of peak lumen stiffness and differences in patient-specific average surface stiffness for homogeneous and element-specific models suggest the role of plaque composition on surface stiffness in addition to local arterial diameter and thickness.

Comparison of 1D and 3D Models for the Estimation of Fractional Flow Reserve

In this work we propose to validate the predictive capabilities of one-dimensional (1D) blood flow models with full three-dimensional (3D) models in the context of patient-specific coronary hemodynamics in hyperemic conditions. Such conditions mimic the state of coronary circulation during the acquisition of the Fractional Flow Reserve (FFR) index. Demonstrating that 1D models accurately reproduce FFR estimates obtained with 3D models has implications in the approach to computationally estimate FFR. To this end, a sample of 20 patients was employed from which 29 3D geometries of arterial trees were constructed, 9 obtained from coronary computed tomography angiography (CCTA) and 20 from intra-vascular ultrasound (IVUS). For each 3D arterial model, a 1D counterpart was generated. The same outflow and inlet pressure boundary conditions were applied to both (3D and 1D) models. In the 1D setting, pressure losses at stenoses and bifurcations were accounted for through specific lumped models. Comparisons between 1D models (FFR1D) and 3D models (FFR3D) were performed in terms of predicted FFR value. Compared to FFR3D, FFR1D resulted with a difference of 0.00 ± 0.03 and overall predictive capability AUC, Acc, Spe, Sen, PPV and NPV of 0.97, 0.98, 0.90, 0.99, 0.82, and 0.99, with an FFR threshold of 0.8. We conclude that inexpensive FFR1D simulations can be reliably used as a surrogate of demanding FFR3D computations.

Coronary fractional flow reserve derived from intravascular ultrasound imaging: Validation of a new computational method of fusion between anatomy and physiology

OBJECTIVES

To evaluate the diagnostic performance of a novel computational algorithm based on three-dimensional intravascular ultrasound (IVUS) imaging in estimating fractional flow reserve (IVUS_FR ), compared to gold-standard invasive measurements (FFR_INVAS ).

BACKGROUND

IVUS provides accurate anatomical evaluation of the lumen and vessel wall and has been validated as a useful tool to guide percutaneous coronary intervention. However, IVUS poorly represents the functional status (i.e., flow-related information) of the imaged vessel.

METHODS

Patients with known or suspected stable coronary disease scheduled for elective cardiac catheterization underwent FFR_INVAS measurement and IVUS imaging in the same procedure to evaluate intermediate lesions. A processing methodology was applied on IVUS to generate a computational mesh condensing the geometric characteristics of the vessel. Computation of IVUS_FR was obtained from patient-level morphological definition of arterial districts and from territory-specific boundary conditions. FFR_INVAS measurements were dichotomized at the 0.80 threshold to define hemodynamically significant lesions.

RESULTS

A total of 24 patients with 34 vessels were analyzed. IVUS_FR significantly correlated (r = 0.79; P < 0.001) and showed good agreement with FFR_INVAS , with a mean difference of -0.008 ± 0.067 (P = 0.47). IVUS_FR presented an overall accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of 91%, 89%, 92%, 80%, and 96%, respectively, to detect significant stenosis.

CONCLUSION

The computational processing of IVUS_FR is a new method that allows the evaluation of the functional significance of coronary stenosis in an accurate way, enriching the anatomical information of grayscale IVUS.

Mechanical Characterization of the Vessel Wall by Data Assimilation of Intravascular Ultrasound Studies

Atherosclerotic plaque rupture and erosion are the most important mechanisms underlying the sudden plaque growth, responsible for acute coronary syndromes and even fatal cardiac events. Advances in the understanding of the culprit plaque structure and composition are already reported in the literature, however, there is still much work to be done toward in-vivo plaque visualization and mechanical characterization to assess plaque stability, patient risk, diagnosis and treatment prognosis. In this work, a methodology for the mechanical characterization of the vessel wall plaque and tissues is proposed based on the combination of intravascular ultrasound (IVUS) imaging processing, data assimilation and continuum mechanics models within a high performance computing (HPC) environment. Initially, the IVUS study is gated to obtain volumes of image sequences corresponding to the vessel of interest at different cardiac phases. These sequences are registered against the sequence of the end-diastolic phase to remove transversal and longitudinal rigid motions prescribed by the moving environment due to the heartbeat. Then, optical flow between the image sequences is computed to obtain the displacement fields of the vessel (each associated to a certain pressure level). The obtained displacement fields are regarded as observations within a data assimilation paradigm, which aims to estimate the material parameters of the tissues within the vessel wall. Specifically, a reduced order unscented Kalman filter is employed, endowed with a forward operator which amounts to address the solution of a hyperelastic solid mechanics model in the finite strain regime taking into account the axially stretched state of the vessel, as well as the effect of internal and external forces acting on the arterial wall. Due to the computational burden, a HPC approach is mandatory. Hence, the data assimilation and computational solid mechanics computations are parallelized at three levels: (i) a Kalman filter level; (ii) a cardiac phase level; and (iii) a mesh partitioning level. To illustrate the capabilities of this novel methodology toward the in-vivo analysis of patient-specific vessel constituents, mechanical material parameters are estimated using in-silico and in-vivo data retrieved from IVUS studies. Limitations and potentials of this approach are exposed and discussed.