0094

Sensitivity Study of 4D flow Left Ventricular Hemodynamics Parameters in Healthy Volunteers and Dilated Cardiomyopathy Patients
Pamela Alejandra Franco1,2,3, Julio Sotelo1,2,3, Bram Ruijsink4, David Nordsletten4, Eric Kerfoot4, Joaquín Mura1,2, Cristian Tejos1,2,3, Daniel Hurtado1,5,6, and Sergio Uribe1,2,6,7

1Millennium Nucleus for Cardiovascular Magnetic Resonance, Santiago, Chile, 2Biomedical Imaging Center, Pontificia Universidad Católica de Chile, Santiago, Chile, 3Electrical Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile, 4Biomedical Engineering, King's College London, London, United Kingdom, 5Structural and Geotechnical Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile, 6Institute for Biological and Medical Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile, 7Radiology Department, Pontificia Universidad Católica de Chile, Santiago, Chile

Synopsis

Segmentation of the Left Ventricle (LV) from MRI and 4D flow datasets is an essential step for quantifying clinical indices and hemodynamic parameters. Automatic methods for heart segmentation are generally validated using cardiac volumetric and global ejection fraction. However, little is known about the changes of hemodynamic parameters when subjected to different segmentations. In this study, we present a sensitivity assessment of left intraventricular hemodynamic parameters in healthy volunteers and Dilated Cardiomyopathy (DCM) patients using a finite element quantification approach using 4D flow MRI data, when subjected to changes of LV segmentations.

Introduction

Understanding the blood flow is essential to understand the physiology and pathophysiology of the cardiovascular system1. Recently, some studies have shown a relationship between the hemodynamics and structural behavior of the heart and may be a sensitive marker of cardiac health2. As an example, Dilated Cardiomyopathy (DCM) patients have a large Left Ventricle (LV). Thus, leading to a high wall tension which causes energy waste3 and altered vortex ring dynamics4, which leads to poor heart pumping efficiency. An important step to define the utility of a cardiovascular parameter is to study its accuracy but also its reproducibility and variability. The variability and reproducibility of functional parameters of the heart such as ejection fraction and stroke volumes has been largely studied by using different LV segmentations strategies6-8,11. However, the sensitivity of new hemodynamic parameters to different LV segmentations have not been assessed thoroughly. The aim of this study is to assess the sensitivity of the LV hemodynamic parameters subjected to different LV segmentations in healthy volunteers and DCM patients using a finite element quantification approach of 4D flow MRI datasets.

Methods

The 4D flow MRI data of 10 healthy volunteers (5 Male, mean age 40.8 ± 9.4 years) and 11 patients (5 Male, mean age 48.5 ± 9.9 year) with DCM were acquired in a clinical MR Scanner of 1.5T (Philips Achieva, Best, The Netherlands). Figure 1 shows the steps for the in-vivo image processing and quantification, which is fully described elsewhere5. Hemodynamic parameters such as velocity, vorticity, helicity, energy loss, and kinetic energy were calculated. LV endocardium was delineated semi–automatically in the short-axis b-SSFP images, using the software Segment v2.2 R6410 (Medviso AB, Lund, Sweden)6-8, for each cardiac phase in the entire cardiac cycle. Thereafter, a sensitivity study was performed by looking at changes of the hemodynamic parameters subjected to changes of the LV segmentation. From the original segmentations, we increased and decreased the size of the LV cavity by varying the location of each point in the contour by 0.5 mm to 2.89 mm, which is equivalent to move the segmentation contour in 1 to 2 pixels. The mean volume and each hemodynamic parameter were compared with the respective mean value of the original LV segmentation, considered as reference. To compare the variables across the different LV segmentation, we use the Kruskall-Wallis Test. The significance level was adjusted by using Dunn’s Test correction. A p-value < 0.05 was considered indicative of statistical significance. The errors and the statistical analysis were measured during peak-systolic, E- wave and end-diastole (see Figure 1.d).

Results

Table 1 shows the clinical data of DCM patients and volunteers. Statistical differences were found in ejection fraction, LV end-diastolic volume, and LV end-systolic volume. The results of our sensitivity analysis at peak-systole, E- wave and end- diastole are shown in Figure 2 to 4. The error in LV cardiac volumes as well as in helicity density showed not significant differences across different LV segmentation. For the other parameters, there were significant differences for some segmentations, particularly for the velocity magnitude, energy loss and kinetic energy, which showed to have larger errors.

Conclusions

We have studied the sensitivity of different hemodynamic parameters in the LV using a FE quantification method. The results showed significant error, particularly for the velocity and those parameters that directly depend on it as the kinetic energy and energy loss. Those parameters that depend on the derivative of the velocity showed less dependency on the segmentation error. In previous studies, it has been shown that DCM patients have a lower mean value of velocity magnitude than healthy patients. This is the case when the segmentation contour falls inside the LV blood pool. In general, we also observed that for almost all hemodynamic parameters lower errors were obtained when the segmentation underestimated the LV volumes than when the segmentation was overestimated.

Acknowledgements

This publication has received funding from Millenium Science Initiative of the Ministry of Economy, Development and Tourism, grant Nucleus for Cardiovascular Magnetic Resonance. Also, has been supported by CONICYT - PIA - Anillo ACT1416, CONICYT FONDEF/I Concurso IDeA en dos etapas ID15|10284, FONDECYT #1181057. Sotelo J. thanks to FONDECYT Postdoctorado 2017 #3170737 and Franco P. thanks to CONICYT – PCHA/ Doctorado-Nacional/2018-21180391.

References

  1. Bleumink GS, et. al., Quantifying the heart failure epidemic: Prevalence, incidence rate, lifetime risk and prognosis of heart failure – The Rotterdam study, Eur. Heart J., vol. 25, pp. 1614-1619, 2004
  2. Föll D, et. al., Age, gender, blood pressure, and ventricular geometry influence normal 3D blood flow characteristics in the left heart, Eur, Heart J., vol. 14, pp. 366-373, 2013.
  3. Al-Wakeel N, et. al., Hemodynamic and energetic aspects of the left ventricle in patients with mitral regurgitation before after mital valve surgery. J Magn Reson Imaging 2015 Dec; 42(6):1705-12.
  4. Arvidsson PM, et. al., Vortex ring behavior provides the epigenetic blueprint for the human heart. Sci Rep 2016; 6 22021.
  5. Franco P, et. al., Comprehensive 3D flow characterization in patients with Dilated Cardiomyopathy from 4D flow MRI data using a finite element method and 17 – Segment Bullseyes, Oral Presentation Velocity & Flow 8:15-10:15, Jun 20, 2018. ISMRM annual meeting, Jun 16-21th, Paris, France http://indexsmart.mirasmart.com/ISMRM2018/PDFfiles/0692.html
  6. Heiberg E, et. al., Design and Validation of Segment – a Freely Available Software for Cardiovascular Image Analysis, BMC Medical Imaging, 10:1, 2010.
  7. Tufvesson J, et. al., Validation and Development of a New Automatic Algorithm for Time-Resolved Segmentation of the Left Ventricle in Magnetic Resonance Imaging, Biomed Res Int, 2015:970357.
  8. Heiberg E, et. al., Time-Resolved Three-dimensional Segmentation of the Left Ventricle, In Proceedings of IEEE Computers In Cardiology 2005(32), pp 599-602, Lyon, France, 2005.
  9. Sotelo J, et. al., 3D Quantification of Wall Shear Stress and Oscillatory Shear Index Using a Finite-Element Method in 3D CINE PC-MRI Data of the Thoracic Aorta. IEEE Trans Med Imaging. 2016 Jun;35(6):1475-87.
  10. Sotelo J, et. al., Three-dimensional quantification of vorticity and helicity from 3D cine PC MRI using finite-element interpolations. Magn Reson Med. 2017;10(1002/mrm.26687).
  11. Teo S-K, et. al., Regional ejection fraction and regional area strain for left ventricular function assessment in male patients after first-time myocardial infarction. J. R. Soc. Interface 12: 20150006. http://dx.doi.org/10.1098/rsif.2015.0006

Figures

Figure 1. Schematic description of the steps in the data analysis. First, we perform the alignment of the 4D Flow data and the 3D bSSFP images (a), 3D bSSFP interpolation used to generate a thinner slice thickness (b), LV segmentation and tetrahedral mesh generation (c), we found specific cardiac times, to decrease the uncertainty, we estimated those values as the mean of adjacent cardiac phases according to the peak studied (d), we computed the velocity vector at each node of the mesh from the 4D flow MRI datasets by using a cubic interpolation, using FE method we calculated HP9,10 (e).

Table 1. Demographical and clinical data for the healthy and DCM patients. All quantitative data are expressed as the mean ± SD.

Figure 2. Error values of the volume (a) and each HP (b-h), for each group of volunteers and patients at peak – systolic. * indicates statistical significant differences (p < 0.05).

Figure 3. Error values of the volume (a) and each HP (b-h), for each group of volunteers and patients at E - wave. * indicates statistical significant differences (p < 0.05).

Figure 4. Error values of the volume (a) and each HP (b-h), for each group of volunteers and patients at E - wave. * indicates statistical significant differences (p < 0.05).

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)
0094