Robust clinical cardiac Echo Particle Image Velocimetry (EchoPIV)

Robust clinical cardiac Echo Particle Image Velocimetry (EchoPIV)

Meyers, B.
Charanko, J.
Pu, M.
Little, W.
Vlachos, P.

2013-07-02

Heart disease is one of the primary causes of morbidity and mortality throughout the world and non-invasive imaging modalities such as echocardiographic ultrasound and magnetic resonance imaging (MRI) are often used for diagnosis or to guide treatment. However, these methods either provide limited insight into fluid dynamics (ultrasound) or are slow and expensive (MRI). Echo Particle Image Velocimetry (Echo-PIV) is a developing diagnostic approach with the potential to overcome limitations of current methods, providing good spatial and temporal resolution information for improved understanding of cardiac flows. Although showing promise, a minority of the current literature has presented analysis of clinical data, and only in a research setting, not routine clinical practice, since use of contrast agents is only clinically indicated for subjects with poor ultrasound image quality. This restriction makes use of Echo-PIV on these patients difficult. To overcome the limitation imposed by image quality in routine clinical scans, we propose the use of short-time moving window ensemble (MWE) methods to improve signal-to-noise (SNR) ratios while preserving the instantaneous fluid mechanics of the results. In addition to traditional summed correlations, we also implemented a product of correlations (PoC) ensemble method to further increase correlation SNR. Tests were conducted on a cohort of 14 patients for whom the use of ultrasound contrast agent was clinically indicated, and the quality of the resultant vector fields was compared between instantaneous cross correlation and the two new ensemble methods. For single frame measurements with standard cross correlation, only 43% of the vectors were likely to be reliable with peak ratios (PPR) over 2.0, while about 90% of both ensemble methods had PPR > 2.0, slightly higher for the PoC. Using frameto- frame velocity difference as a proxy for random error, the standard deviation of velocity difference was at least twice as large for the instantaneous correlations as for MWE. Qualitative examination of the flow fields showed vectors fields that were more physically plausible, with fewer areas of apparent loss of correlation and better agreement with color M-mode scans (CMM). Finally, in comparison to clinically measured inflow velocities from mitral inflow Doppler and CMM ultrasound, fits of echo-PIV values showed lower RMS error and higher R2 values for MWE than for the single-frame correlations. In conclusion, moving window ensemble in conjunction with a product of correlation method demonstrates the ability to provide credible velocity information throughout the heart cycle even in routine clinical scans with low SNR for which traditional instantaneous correlation methods largely fail to resolve the flow physics.

http://resolver.tudelft.nl/uuid:a877f3fb-f629-438a-847d-424418e8b0dd

Conference proceedings

conference paper

(c) 2013 Meyers, B.; Charanko, J.; Pu, M.; Little, W.; Vlachos, P.