Contrast echocardiography for cardiac quantifications

Research output: ThesisPhd Thesis 1 (Research TU/e / Graduation TU/e)

581 Downloads (Pure)

Abstract

The indicator-dilution-theory for cardiac quantifications has always been limited in practice by the invasiveness of the available techniques. However, the recent introduction of stable ultrasound contrast agents opens new possibilities for indicator dilution measurements. This study describes a new and successful approach to overcome this invasiveness issue. We show a novel approach for minimally invasive quantification of several cardiac parameters based on the dilution of ultrasound contrast agents. A single peripheral injection of an ultrasound contrast agent bolus can result in the simultaneous assessment of cardiac output, pulmonary blood volume, and left and right ventricular ejection fraction. The bolus passage in different sites of the central circulation is detected by an ultrasound transducer. The detected acoustic (or video) intensities are processed and several indicator dilution curves are measured simultaneously. To this end, we exploit that for low concentrations the relation between contrast concentration and acoustic backscatter is approximately linear. The Local Density Random Walk Model is used to fit and interpret the indicator dilution curves for cardiac output, pulmonary blood volume, and ejection fraction measurements. Two fitting algorithms based either on a multiple linear regression in the logarithmic domain or on the solution of the moment equations are developed. The indicator dilution system can be also interpreted as a linear system and, therefore, characterized by an impulse response function. An adaptive Wiener deconvolution filter is implemented for robust dilution system identification. For ejection fraction measurements, the atrial and ventricular indicator dilution curves are measured and processed by the deconvolution filter, resulting in the estimate of the left ventricle dilution-system impulse response. This curve can be fitted and interpreted by a mono-compartment exponential model for the ejection fraction assessment. The proposed deconvolution filter is also used for the identification of the dilution system between right ventricle and left atrium. The Local Density Random Walk Model fit of the estimated impulse response allows the pulmonary blood volume assessment. Both cardiac output and pulmonary blood volume measurements are validated in vitro with accurate results (correlation coefficients larger than 0.99). The Pulmonary blood volume measurement feasibility is also tested in humans with promising results. The ejection fraction measurement is validated in-vivo. The impulse response approach allows accurate left ventricle ejection fraction estimates. Comparison with echocardiographic bi-plane measurements shows a correlation coefficient equal to 0.93. A dedicated image segmentation algorithm for videodensitometry has also been developed for automating the determination of regions of interest. The resulting algorithm has been integrated with the indicator dilution analysis system. The automatic determination of the measurement region results in improved dilution-curve signal-to-noise ratios. In conclusion, this study proves that quantification of cardiac output, pulmonary blood volume, and left and right ventricular ejection fraction by dilution of ultrasound contrast agents is feasible and accurate. Moreover, the proposed methods are applicable in different contexts (e.g., magnetic resonance imaging) and for different types of measurements, leading to a broad range of applications.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Department of Electrical Engineering
Supervisors/Advisors
  • Kalker, Ton, Promotor
  • Korsten, H.H.M (Erik), Promotor
Award date4 Oct 2004
Place of PublicationEindhoven
Publisher
Print ISBNs90-386-1613-9
DOIs
Publication statusPublished - 2004

Fingerprint Dive into the research topics of 'Contrast echocardiography for cardiac quantifications'. Together they form a unique fingerprint.

Cite this