Elsevier

Medical Image Analysis

Volume 38, May 2017, Pages 117-132
Medical Image Analysis

Extraction of the 3D local orientation of myocytes in human cardiac tissue using X-ray phase-contrast micro-tomography and multi-scale analysis

https://doi.org/10.1016/j.media.2017.02.006Get rights and content

Highlights

  • 3D imaging of fresh human cardiac tissue samples using X-ray phase-contrast μCT (ESRF).

  • Each reconstructed sample is 40 Gbytes with an isotropic spatial resolution of 3.5 µm.

  • We propose a multiscale method to extract the 3D arrangement of the oriented inner structures of the cardiac wall at various resolution levels.

  • We demonstrate an increase of the transmural helix angle range estimation with the 3D spatial resolution.

Abstract

This paper presents a methodology to access the 3D local myocyte arrangements in fresh human post-mortem heart samples. We investigated the cardiac micro-structure at a high and isotropic resolution of 3.5 µm in three dimensions using X-ray phase micro-tomography at the European Synchrotron Radiation Facility. We then processed the reconstructed volumes to extract the 3D local orientation of the myocytes using a multi-scale approach with no segmentation. We created a simplified 3D model of tissue sample made of simulated myocytes with known size and orientations, to evaluate our orientation extraction method. Afterwards, we applied it to 2D histological cuts and to eight 3D left ventricular (LV) cardiac tissue samples. Then, the variation of the helix angles, from the endocardium to the epicardium, was computed at several spatial resolutions ranging from 3.63 mm3 to 1123 µm3. We measure an increased range of 20° to 30° from the coarsest resolution level to the finest level in the experimental samples. This result is in line with the higher values measured from histology. The displayed tractography demonstrates a rather smooth evolution of the transmural helix angle in six LV samples and a sudden discontinuity of the helix angle in two septum samples. These measurements bring a new vision of the human heart architecture from macro- to micro-scale.

Introduction

Cardiovascular diseases remain one of today’s most serious health problems, which motivates the search for a deeper understanding of myocardial function. The relation between mechanical function, hemodynamics, perfusion, diffusion, transfer rate and the adaptive structural changes emerging in cardiac diseases remain poorly understood. In addition, a better understanding of the anatomy of 3D tissue would be a great help to further the understanding of cardiac pathophysiology. This would require producing a 3D multi-scale atlas of the human heart from the cell level to that of the organ, unavailable today.

Cardiac myocytes are approximately cylindrical, 10 µm in diameter and 100 µm long. Their membrane is about 8 nm thick (Iaizzo, 2009). Today, it is not possible to image a whole adult human heart at a micrometric resolution (<3 µm), thereby providing the individual myocyte arrangement (for the sake of simplification, we refer to these fibres as myocytes arrangements) within the tissue. Histological slices deliver high-resolution (µm) 2D images of ex vivo samples. Polarised light imaging (PLI, Jouk et al., 2007) provides the 3D orientation of populations of myocytes (100 × 100 ×500 µm3) of ex vivo hearts based on their birefringence property. Magnetic resonance cardiac diffusion tensor imaging (cDTI) measures the magnitude and direction of intra-myocardial water diffusion (Ferreira et al., 2014) (2.1 × 1.7 × 6 mm3), but various assumptions on the tissue’s micro-architecture (Frindel, Robini, Croisille, Zhu, 2009, Ferreira, Kilner, McGill, Nielles-Vallespin, Scott, Ho, McCarthy, Haba, Ismail, Gatehouse, de Silva, Lyon, Prasad, Firmin, Pennell, 2014, Bernus, Radjenovic, Trew, LeGrice, Sands, Magee, Smaill, Gilbert, 2015) are still necessary to interpret in vivo measurements and cardiac muscle mobility in normal and diseased hearts (Wang et al., 2012). High-resolution MRI setups (Baltes et al., 2009) can achieve 30 × 30 × 300 µm3, but this is larger than the size of myocytes. At the same time, μCT imaging has been used by Stephenson et al. (2012) to image rat and rabbit hearts at a resolution within the 13–41 µm range to estimate heart morphology. Similarly, van Deel et al. (2016) imaged in vivo mice hearts to extract quantitative metabolic markers at an isotropic 40 µm resolution. Recently, some preliminary fibre analysis has been conducted in the atrial section of canine hearts by Aslanidi et al. (2013) at an isotropic 36 µm resolution and in fresh ex vivo human hearts by Zhao et al. (2015) at an isotropic resolution of 49 µm. However, these techniques are only based on X-ray absorption and they all use an isotropic resolution, which is higher than the lateral dimensions of a cardiomyocyte. A review of the literature examining studies dedicated to the measurement of the local 3D cardiomyocyte orientation was conducted (Appendix). It was restricted to studies providing measurements of the helix angle range through the cardiac wall. It can be noted that this range differs with the imaging modality, the protocol used and the heart model (mouse, rat, dog, human). However, maximum values are observed when very high spatial resolution techniques are used (Streeter, Bassett, 1966, Streeter, Spotnitz, Patel, Ross, Sonnenblick, 1969, LeGrice, Smaill, Chai, Edgar, Gavin, Hunter, 1995, Gilbert, D., A.P., White, Tanner, A.V., Dobrzynski, Bernus, Radjenovic, 2012).

In this study, our aim was to access the local 3D myocyte arrangements within fresh human post-mortem heart samples. To handle the above-mentioned limitations, we used the synchrotron radiation phase micro-computed tomography setup (SR-PCT) at the European Synchrotron Radiation Facility (ESRF) beamline ID19 to image the samples using phase contrast imaging, at a voxel size of 3.5 µm. X-ray phase contrast imaging yields several orders of magnitude higher sensitivity in light materials compared to standard X-ray attenuation-based imaging (Momose et al., 1995). Phase contrast SR-PCT is increasingly used to characterize soft tissue due to its histology-like contrast (Pierre, Langer, Boistel, Cloetens, 2007, Boistel, Aubin, Cloetens, Langer, Gillet, Josset, Pollet, Herrel, 2011, Marinescu, Langer, Durand, Olivier, Chabrol, Rositi, Chauveau, Cho, Nighoghossian, Berthezène, Peyrin, Wiart, 2013, Lang, Zanette, Dominietto, Langer, Rack, Schulz, Le Duc, David, Mohr, Pfeiffer, Muller, Weitkamp, 2014). The first acquisitions were made on a healthy fetal rabbit heart (Baličević et al., 2015) and a healthy human heart (Varray et al., 2013). The isotropic 3D spatial resolution of SR-PCT makes it possible to represent a myocyte by several voxels and to process the data to estimate the transmural myocyte arrangements (in terms of orientation) at several scales. The reconstructed tomographic images were previously compared to corresponding histological images, to confirm that the cardiomyocytes were visible (Mirea et al., 2015).

In this paper, the 3D-acquired tomograms were processed to extract the local fibre arrangement by measuring the transmural (helix and transverse) fibre angles. We developed a multi-scale scheme to extract the orientation of the fibres at various spatial resolutions and to evaluate the accuracy of the angular estimation. Finally, we used a fibre tracking technique to display the fibres arrangements of several cardiac tissue samples.

The paper is organised as follows. First, we present the cardiac tissue samples and the acquisition procedure. Second, the multi-scale method used to extract the 3D cardiomyocyte arrangements is described. Third, the results obtained at different resolutions in both simulations and experiments are highlighted. A discussion concludes the paper.

Section snippets

Extraction and preparation of the human heart samples

Eight fresh human heart samples were supplied by the Medico-Legal Institute of Lyon IML HCL (N° DC-2012-1588), taken from the hearts of two males aged 32 and 35 years. These patients died a violent death and had no past medical history, no treatment and no addiction. Less than 24 h after death, the hearts were removed and preserved in a 10% formalin solution to ensure good preservation. A slight shrinkage effect due to the formalin solution was present but limited by the short time between

Extraction of the local orientation of myocytes

In SR-PCT images, the orientation of myocytes is directly related to the fibres’ local orientation. We propose a dedicated methodology to extract this 3D orientation and compute both θ and ϕ values at the same time. We first computed the module of the 3D Fourier transform (FT) of one 3D sub-volume Vsub of the reconstructed volume: V^=|FT(Vsub)|

In the spatial domain, the stack of myocytes has a particular arrangement in beams or lines or “fibres” (Fig. 7a). In the Fourier domain (Fig. 7b), such

Method validation

We tested our local fibres orientation extraction method using 2D histological heart tissue sections extracted immediately in the vicinity of the LV posterior sample (Fig. 9). We also selected two image sections extracted from the 3D X-ray reconstructed volumes of sample #6 (LV_post_S1_2), one with homogeneous tissue (mainly myocytes) and one containing other biological material (vessel, cleavage planes, etc.). We computed the orientation of the myocytes, on each 2D image using our method. We

Discussion

In this paper, the fibres structure of the human heart is investigated at a high resolution using both simulated and real data. The data are X-ray SR-PCT images of eight LV cardiac tissue samples, acquired at ID19, ESRF, and reconstructed at an isotropic voxel size of 3.5 µm. By extracting the fibres orientation in the Fourier domain, the low-pass shape of the myocyte arrangements was identified and gave the local direction of the fibres through the LV wall. The myocyte orientation was

Acknowledgements

The authors thank Cécile Olivier, Lihui Wang, David Rousseau, and Matthieu Ozon for their help for acquiring data at the ESRF and Pierre Croisille and Yue-Min Zhu for fruitful discussions. They also thank the ESRF for support throughout the MD-701 experiment and for the assistance of Alexander Rack and Anne Bonnin from the ESRF ID19 group. This work was done within the framework of LabEx PRIMES (ANR-11-LABX-006).

References (40)

  • A. Brüel et al.

    Design–based stereological estimation of the total number of cardiac myocytes in histological sections

    Basic Res. Cardiol.

    (2005)
  • E. van Deel et al.

    In vivo quantitative assessment of myocardial structure, function, perfusion and viability using cardiac micro-computed tomography

    J. Visual. Exp.: JoVE

    (2016)
  • R. Dettmeyer

    Forensic Histopathology: Fundamentals and Perspectives

    (2011)
  • L. Feldkamp et al.

    Practical cone-beam algorithm

    J. Optical Soc. Am. A

    (1984)
  • S. Gilbert et al.

    Visualization and quantification of whole rat heart laminar structure using high-spatial resolution contrast-enhanced mri

    Am. J. Physiol. - Heart Circulat. Physiol.

    (2012)
  • R.A. Greenbaum et al.

    Left ventricular fibre architecture in man

    Br. Heart J.

    (1981)
  • P.A. Helm et al.

    Ex vivo 3d diffusion tensor imaging and quantification of cardiac laminar structure

    Magn. Reson. Med.

    (2005)
  • P.A. Iaizzo

    Handbook of Cardiac Anatomy, Physiology, and Devices

    (2009)
  • P.-S. Jouk et al.

    Analysis of the fiber architecture of the heart by quantitative polarized light microscopy. accuracy, limitations and contribution to the study of the fiber architecture of the ventricles during fetal and neonatal life.

    Eur. J. Cardiothorac. Surg.

    (2007)
  • S. Lang et al.

    Experimental comparison of grating- and propagation-based hard x-ray phase tomography of soft tissue

    J. Appl. Phys.

    (2014)
  • Cited by (29)

    • Measurement of local orientation of cardiomyocyte aggregates in human left ventricle free wall samples using X-ray phase-contrast microtomography

      2022, Medical Image Analysis
      Citation Excerpt :

      The transmural distributions of the orientation of myocyte aggregates calculated by the classic method and the skeleton method are very similar. These results confirm the preliminary results of Varray et al. (2017), and show that at a global scale, the behaviors of the two methods are comparable. However, even if the skeleton method can provide angular values at a global scale, it needs a preliminary binarization step and it is much more time consuming than the classic method, so it is not designed for this.

    • Imaging biomarkers for cardiovascular diseases

      2019, Handbook of Medical Image Computing and Computer Assisted Intervention
    • Description of the Intrusion Angle of Local Cardiomyocyte Aggregates in Human Left Ventricular Free Wall Using X-ray Phase-Contrast Tomography

      2023, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    View all citing articles on Scopus
    View full text