Original ContributionDevelopment and Validation of a Phase-Filtered Moving Ensemble Correlation for Echocardiographic Particle Image Velocimetry
Introduction
Recent advancements in cardiac flow visualization have changed the way cardiac disease progression is studied (del Álamo et al, 2009, Hong et al, 2013, Muñoz et al, 2013). Ultrasound and magnetic resonance imaging (MRI) now provide high-resolution information on blood flow within the chambers of the heart. Despite advancement, inherent limitations for each modality still exist. Color M-mode (CMM) collects blood velocity measurements along a thin scan line, preventing visualization of flow structures, whereas 2-D Color Doppler has low temporal resolution (Muñoz et al. 2013). Further, color Doppler modalities are accurate only at high velocities and unreliable for low-velocity measurements. MRI velocity measurements also provide low temporal resolution, and high operational costs limit widespread use (Muñoz et al, 2013, Sengupta et al, 2012).
Echo particle image velocimetry (EchoPIV) is an emerging method that provides 2-D planar views with temporal resolution higher than that of color Doppler and MRI, which allows improved visualization of cardiac flow (Muñoz et al, 2013, Sengupta et al, 2012). EchoPIV employs contrast-enhanced B-mode ultrasound imaging, commonly used in cardiac chamber geometry measurements. Flow measurements are made through Fourier-based correlation algorithms commonly used in image processing (Willert and Gharib 1991). Applications of EchoPIV range from cylindrical flow through peripheral arteries and veins (Kim et al, 2004, Zhang et al, 2011) to complex flow through the heart (Sengupta et al. 2007). General development and limitations of EchoPIV are provided in review articles (Bazilevs et al, 2010, Bermejo et al, 2015, Borazjani et al, 2013, del Álamo et al, 2009, Hong et al, 2013, Muñoz et al, 2013, Pedrizzetti, Domenichini, 2015, Poelma, 2017, Sengupta et al, 2012).
Clinical EchoPIV research studies have focused primarily on left ventricular (LV) flows, analyzing the vortexes that form during diastolic filling, linking properties of the vortex to cardiac health (Gharib et al. 2006), although agreement on universal relationships varies (Ghosh et al, 2009, Stewart et al, 2012). Complex topology of flow within the left ventricle presents a challenging condition for many non-invasive imaging modalities because of the need for adequate frame rates and spatial resolution (Arvidsson et al, 2016, Hendabadi et al, 2013). This has motivated use of EchoPIV as one of the primary tools for the study of blood flow within the left ventricle and has allowed researchers to investigate cohorts that include healthy patients (Cimino et al. 2012), LV diastolic dysfunction (LVDD) (Prinz et al. 2013), heart failure (Abe et al. 2013) and congenital heart disease (Kutty et al, 2014, Lampropoulos et al, 2012). Each of these works has furthered the potential diagnostic capabilities of vortex properties.
Although promising, these studies employ optimized scan settings and well-controlled contrast agent infusion (Gao et al. 2011) for performing EchoPIV flow visualization, which is not always achievable in standard clinical settings. Furthermore, the majority of these studies forego capture of accurate peak velocities and, instead, obtain the “general” flow pattern, biasing quantitative measurements of vortex rotation and circulation. There is a need to develop robust EchoPIV methodologies, techniques that consistently provide accurate velocity measurements with limited or no erroneous measurements, in the presence of increasing and spatiotemporally varying noise and image artifacts as they appear in EchoPIV images. With velocity measurements that are accurate in magnitude and direction, clinical measures of diastolic function such as propagation velocity (Vp) and intraventricular pressure difference (IVPD), along with complex flow observations (i.e., vortex shape and position), can be obtained and related from a single scan.
In the work described here, we implemented and tested a short-time moving ensemble (ME) correlation on contrast-enhanced B-mode echocardiograms, coupled with dynamic phase filtering of the cross-correlation, and tested the hypothesis that this approach results in more robust, high-fidelity EchoPIV velocity measurements. The proposed method was validated using artificial ultrasound images, and subsequently we determined its feasibility using a cohort of patients with clinically diagnosed LVDD. For each subject, the resulting EchoPIV velocity measurements were compared against respective CMM velocity measurements. Further time series analysis was conducted to determine ME EchoPIV capabilities for measuring filling wave Vp and IVPD.
Section snippets
Standard pairwise cross-correlation
Particle image velocimetry uses cross-correlation between two sequential images to estimate particle image pattern displacements; herein we refer to this standard approach as pairwise correlation (Willert and Gharib 1991). The method was developed for images obtained from laser-illuminated particle-seeded flow fields that, under proper seeding, illumination, and exposure, produce bright-intensity particle images on dark, low-intensity backgrounds. Contrast-enhanced B-mode images taken for
Artificial data: error analysis
Error measurements from artificial image analysis (Fig. 5) indicate improvement in measurement robustness based on the proposed ME processing algorithm. Pairwise correlation bias error measurements indicate that traditional processing has the tendency to underpredict small displacements compared with the proposed ME correlation (βPairwise = −0.1526 ppf, βME = −0.0624 ppf). Furthermore, random error measurements revealed an improvement in measurement precision for the proposed methodology
Discussion
We have described a short time-series ME PIV with a novel phase filter that obtains more accurate measurements for clinical cardiac EchoPIV. To the authors' knowledge this is the first application of phase-filtered correlation in EchoPIV. Our novel phase filter utilizes the particle sizes in each interrogation region, which are affected by local image magnification, to design a more optimal filter. Eckstein, Vlachos, 2009a, Eckstein, Vlachos, 2009b introduced what is known as robust phase
Conclusions
We have described the application of short time-series moving ensemble PIV with phase-filtered correlation that is better suited for clinical cardiac EchoPIV. Error analysis reveals a twofold increase in accuracy on representative low-quality echocardiogram images. Application to clinical data indicates that ME EchoPIV can better resolve peak inflow and provide better overall agreement to CMM scans compared with traditional EchoPIV processing. We also described, for the first time,
Acknowledgments
This material is based on work supported by the National Institutes of Health [R21 Grant No. HL106276-01 A1].
References (47)
- et al.
From imaging to prediction: Emerging non-invasive methods in pediatric cardiology
Prog Pediatr Cardiol
(2010) - et al.
In vivo analysis of intraventricular fluid dynamics in healthy hearts
Eur J Mech B Fluids
(2012) - et al.
Recent advances in the application of computational mechanics to the diagnosis and treatment of cardiovascular disease
Rev Esp Cardiol
(2009) - et al.
Ensemble correlation PIV applied to bubble plumes rising in a bubble column
Chem Eng Sci
(1999) - et al.
Combined experimental and numerical analysis of the flow structure into the left ventricle
J Biomech
(2007) - et al.
Myocardial strain rate is a superior method for evaluation of left ventricular subendocardial function compared with tissue Doppler imaging
J Am Coll Cardiol
(2003) - et al.
Characterization and quantification of vortex flow in the human left ventricle by contrast echocardiography using vector particle image velocimetry
JACC Cardiovasc Imaging
(2008) - et al.
Effects of right ventricular hemodynamic burden on intraventricular flow in tetralogy of Fallot: An echocardiographic contrast particle imaging velocimetry study
J Am Soc Echocardiogr
(2014) - et al.
Left ventricular isovolumic flow sequence during sinus and paced rhythms: New insights from use of high-resolution Doppler and ultrasonic digital particle imaging velocimetry
J Am Coll Cardiol
(2007) - et al.
Emerging trends in CV flow visualization
JACC Cardiovasc Imaging
(2012)
Evaluation of LV diastolic function from color M-mode echocardiography
JACC Cardiovasc Imaging
In vitro and preliminary in vivo validation of echo particle image velocimetry in carotid vascular imaging
Ultrasound Med Biol
Contrast echocardiography for assessing left ventricular vortex strength in heart failure: A prospective cohort study
Eur Heart J Cardiovasc Imaging
Vortex ring behavior provides the epigenetic blueprint for the human heart
Sci Rep
The clinical assessment of intraventricular flows
Annu Rev Fluid Mech
Left ventricular flow analysis: Recent advances in numerical methods and applications in cardiac ultrasound
Comput Math Methods Med
Vortices formed on the mitral valve tips aid normal left ventricular filling
Ann Biomed Eng
Assessment of advanced windowing techniques for digital particle image velocimetry (DPIV)
Meas Sci Technol
Digital particle image velocimetry (DPIV) robust phase correlation
Meas Sci Technol
Phase correlation processing for DPIV measurements
Exp Fluids
Improvements on the accuracy of derivative estimation from DPIV velocity measurements
Exp Fluids
How to optimize intracardiac blood flow tracking by echocardiographic particle image velocimetry? Exploring the influence of data acquisition using computer-generated data sets
Eur Heart J Cardiovasc Imaging
Optimal vortex formation as an index of cardiac health
Proc Natl Acad Sci
Cited by (6)
Time-resolved particle image velocimetry
2021, Measurement Science and TechnologyColour-Doppler echocardiography flow field velocity reconstruction using a streamfunction-vorticity formulation
2020, Journal of the Royal Society InterfaceExperimental Investigation of the Effect of Heart Rate on Flow in the Left Ventricle in Health and Disease - Aortic Valve Regurgitation
2020, Journal of Biomechanical Engineering
Conflicts of Interest: The authors have no conflicts of interest to report.