Remote analysis of myocardial fiber information in vivo assisted by cloud computing

https://doi.org/10.1016/j.future.2018.03.019Get rights and content

Highlights

  • A novel automatic approach to detect the bubble-like frames in capsule video endoscopy is presented, which is based on a new ring shape selective filter.

  • Matrix eigen values are used to construct the ring selective filters to enhance the boundary areas of the bubbles.

  • The bubble-like frames can be distinguished by the pixel number of the bright area in the filtered images.

Abstract

A cloud-based analysis system of cardiac images is constructed to realize the remote sharing of cardiac images and related computing services. In its core service, a novel image post-processing approach is presented to obtain information on individual in vivo myocardial fibers. This approach is based on the anisotropic decomposition of the local deformation using tagged magnetic resonance (tMR) images of the heart. The local sine wave model (SinMod) is used in our approach to trace the motion in image sequences. Within this motion framework, a pair of anisotropic deformation components for each pixel is then extracted by Poisson orthogonal composition, which can represent exactly the local Poisson Effect. Based on these pair components, the direction structure (by the major deformation vector field) and the elastic property (by Poisson ratio) of myocardial fiber can be estimated. The experimental results demonstrate that our proposed approach is useful in analyzing information of in vivo myocardial fibers. It will have potential to provide the valuable cloud computing service for remote cardiac diagnosis and treatment.

Introduction

In recent years, heart disease has become one of the top leading causes of death in urban and rural population [[1], [2], [3]]. Study on the essential disciplines of cardiac physiology and pathology is necessary to the diagnosis and treatment of heart disease. The myocardial fiber of the heart has a typical anisotropic architecture that is known to play a critical role in determining the cardiac mechanical and conduction properties. Discovering its anisotropic architecture, especially the fiber direction, is the most important foundation work in cardiology [[4], [5], [6]]. However, because of the limitations of medical imaging techniques, the acquisition and reconstruction of myocardial fiber information has been challenging.

In our study, a novel post-processing approach is proposed on tagged magnetic resonance (tMR) images of the heart to analyze the myocardial fiber information in vivo concealed in myocardium deformation. Moreover, in order to acquisite these information expediently and quickly in anywhere as they are needed, a cloud-based analysis system [[7], [8], [9]] of cardiac images is constructed to try to realize the remote analysis service of myocardial fiber information. The following introduction is to derive the solution strategy of this analysis service, which is a different way from the existed cardiac imaging techniques.

At present, diffusion tensor magnetic resonance (DT-MR) is a noninvasive medical imaging technology to exclusively obtain accurate information on tissue fibers [[10], [11]]. This method tries to measure the diffusion properties of water in biological tissue. Regarding the anisotropy in the myocardial fiber architecture, water diffusion along the long axis of the fiber is the fastest, which coincides with the primary eigenvector of the diffusion tensor in DT-MR images [12]. However, breathing and cardiac beating motion often lead to serious signal loss and imaging distortion under the lengthy scanning time for DT-MR images, which is a major deficiency that is not easily overcome [[13], [14]]. Although several studies on this in vivo technology have been started, its practicability has remained relatively limitations [[14], [15], [16]]. Therefore, the personalized architecture of myocardial fibers is still difficult to obtain in vivo from cardiac DT-MR images [17].

To resolve these problems, some researchers tried to fuse complementary information from multi-modal images of the heart. In these approaches, a template of the myocardial fiber architecture is constructed ex vivo by using DT-MR images from a cardiac atlas. At the same time, the geometric information of the specific heart is in vivo extracted by other imaging technologies such as computed tomography (CT) [10] or tMR [18]. The myocardial fiber direction of the atlas is then mapped onto the geometric shape of the specific myocardium using elastic registration and space transformation technology so that the fiber local direction of the specific heart can be estimated. Although these methods may be able to reconstruct the local structure of personalized myocardial fibers in vivo, the results are not accurate enough [[19], [20]]. The main deficiency is that the atlas and the specific heart are from different subjects, and estimation error is introduced by the differences between individuals.

To avoid this deficiency, some researchers have started to explore the reconstruction of personalized myocardial fibers in vivo from a new perspective. In their research, myocardial anisotropic information other than the diffusion properties of water has been revealed. For instance, Lee [[21], [22]] assessed the myocardial anisotropy using supersonic shear wave imaging (SSI) and elastic tensor imaging (ETI), which form an ultrasound-based shear wave technique for mapping the mechanical properties of soft tissues at ultrafast frame rates. In their studies, the local direction of the myocardial fiber in the systole was dynamically evaluated in vivo using the elastic anisotropy measured by SSI and ETI. However, in some cases, echocardiography is not considered to be an appropriate technology to supply precise cardiac information due to its non-repeatable operation and imaging range limitation.

Following the myocardial anisotropy analysis, we found some relevant studies showing that the anisotropic structure of the myocardial fiber is also closely associated with the myocardium anisotropic deformation during cardiac beating [[23], [24]]. In a previous study, myofibrils parallel to each other arrange along the long axis of the myocardial cells at the micro level [25]. Due to this anisotropic structure, the deformation at the macro level also appears anisotropic. When a single fiber expands, there is an intense expanding deformation along its long axis, which is always accompanied by a notable contracting deformation along its cross axis (orthogonal to the long axis), and its thickness simultaneously becomes thinner. A similar phenomenon appears in its contracting process. As a single fiber contracts, an intense contracting deformation along its long axis leads to its length being shortened. Simultaneously, a notable expanding deformation occurs along the cross axis of the fiber, which causes its thickness to increase. This phenomenon is called the Poisson effect [26] (This effect is shown in Fig. 1), and it appears as the anisotropic deformation of a myocardial fiber on two perpendicular axes. Similar to the anisotropic diffusion of water molecules in fiber tissue, it can indicate relative information about the myocardial fiber structure. Due to its dynamical properties, it might provide other information of the fiber as well, such as the elasticity. Tagged magnetic resonance (tMR) imaging is a popular technology used in the clinic to obtain high-quality dynamical images of the heart in vivo. It has been used to trace cardiac motion for myocardium deformation analysis [[27], [28]] in previous studies and thus provides a good foundation for our research.

Based on the above survey, a novel approach aimed at analysis and remote acquisition of an in vivo myocardial fiber is presented. It is based on the anisotropic decomposition of the myocardium deformation on tMR images of the heart, which is the key strategy to supply the cloud computing service in our medical image cloud system. The anisotropic feature in the myocardium deformation appeared in the Poisson effect was used to analysis the myocardial fiber information. It is a significative method to expediently and preliminarily know the information about myocardial fiber in routine examinations. In this key strategy, tMR images from a beating heart are firstly employed to trace the myocardium motion by using the SinMod method [[27], [28]]. Then, the relative deformation in the local myocardium is extracted from its traced original motion, which appears notable anisotropy. Subsequently, a pair of anisotropic deformation components is decomposed from this local deformation through Poisson orthogonal decomposition for each pixel. Based on the information contained in this pair, we try to analyze an in vivo myocardial fiber to derive more specific information such as its direction and elastic property. These fiber information can supply to the remote medical experts by cloud to aided the diagnosis and treatment of heart disease. More details and an analysis will be given in the following sections.

The rest of the paper is organized as follows: Section 2 presents the frame of the cloud-based analysis system of in vivo myocardial fiber on cardiac tMR images. In Section 3, the core cloud computing service is explained in detail, which is focused on how to use the Poisson orthogonal decomposition of the deformation on cardiac tMR images to analysis in vivo myocardial fiber information. In Section 4, performance of the proposed scheme in core cloud computing service is evaluated with numerical experiments. The Section 5 summarizes and concludes this paper.

Section snippets

Frame of a cloud-based analysis system on cardiac images

A cloud-based system of cardiac tMR images is constructed to analysis information of in vivo myocardial fiber in our study. The schematic diagram is shown in Fig. 2.

First of all, the cardiac tMR images are collected during the routine examination of heart disease in MR center of different hospitals. Then, these tMR image data are transmitted to the cloud through high speed networks.

Cardiac tMR images are time series images with multi sections, which leads to the relative big image data. It is

Realization of the core cloud service

In this cloud system, analysis of in vivo myocardial fiber information is the crucial approach to provide the core cloud service, which will be described in detail as following. Since this analyzation approach is a quite different way from other existed methods, its explanation will be divided into two parts. At first part (in 3.1), the ideal models of elastic fiber have been constructed to attempt preliminary analysis, which can explain the relationship between the deformation and fiber

Dataset

Two different image datasets are used in our research, provided from GE 1.5T HDxt or Siemens 3.0T skyra MR scanners in our collaboration hospital. The first dataset is used to analyze the myocardial fiber structure. This dataset includes both tMR and DT-MR multi-module images collected from the same dog heart in vivo and ex vivo, respectively. The second dataset is used to analyze the elasticity of the myocardial fiber and includes 21 cases of tMR images of the heart collected in vivo from

Conclusion and future work

Cardiac images and related computing services can be accessed to the remote terminal in our cloud system. The core service can supply information of in vivo myocardial fiber on cardiac tMR Images by a post image processing. In this service, the proposed approach in cardiac tMR images can compute the structure and elastic properties to aid in myocardial fiber analysis in vivo. This approach is based on the anisotropic analysis of myocardium deformation, Poisson orthogonal decomposition of the

Acknowledgments

We would like to thank Tongji Hospital affiliated with the Huazhong University of Science and Technology for providing the experimental datasets in this study and providing useful medical suggestions.

This research was funded by the National Natural Science Foundation of China (Grant No. 61602519); the China Postdoctoral Science Foundation, China (Grant No. 2014M552045); the China Postdoctoral Science Foundation, China (Grant No. 2014M562026); the Natural Science Foundation of Hubei Province,

Qian Wang received her B.S. degree from the Department of Computer Science and Technology, Central China Normal University, Wuhan, China, in 2005, and her Ph.D. degree from the School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan, China, in 2009. She is currently an associate professor with the School of Information and Safety Engineering, Zhongnan University of Economics and Law, Wuhan, China. Her current research interests include computer vision,

References (40)

  • MerchantS.S. et al.

    Parametric modeling of the mouse left ventricular myocardial fiber structure

    Ann. Biomed. Eng.

    (2016)
  • PashakhanlooF.

    Submillimeter diffusion tensor imaging and late gadolinium enhancement cardiovascular magnetic resonance of chronic myocardial infarction

    J. Cardiovasc. Magn. Reson.

    (2017)
  • HossainM.S. et al.

    Smart healthcare monitoring: A voice pathology detection paradigm for smart cities

    Multimedia Syst.

    (2008)
  • HossainM.S.

    Cloud-supported cyber-physical localization framework for patients monitoring

    IEEE Syst. J.

    (2017)
  • M.S. Hossain, G. Muhammad, Emotion-aware connected healthcare big data towards 5G, Multimedia Syst. PP...
  • VadakkumpadanF. et al.

    Image-based estimation of ventricular fiber orientations for personalized modeling of cardiac electrophysiology

    Med. Imaging IEEE Trans.

    (2012)
  • FroelingM. et al.

    Diffusion tensor MRI of the heart-in vivo imaging of myocardial fiber architecture

    Curr. Cardiovasc. Imaging Rep.

    (2014)
  • AquinoC. et al.

    Magnetic resonance imaging of traumatic brain injury: A pictorial review

    Emerg. Radiol.

    (2015)
  • AxelL. et al.

    Probing dynamic myocardial microstructure with cardiac magnetic resonance diffusion tensor imaging

    J. Cardiovasc. Magn. Reson.

    (2015)
  • LauA.Z. et al.

    Accelerated human cardiac diffusion tensor imaging using simultaneous multislice imaging

    Magn. Reson. Med.

    (2015)
  • Cited by (0)

    Qian Wang received her B.S. degree from the Department of Computer Science and Technology, Central China Normal University, Wuhan, China, in 2005, and her Ph.D. degree from the School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan, China, in 2009. She is currently an associate professor with the School of Information and Safety Engineering, Zhongnan University of Economics and Law, Wuhan, China. Her current research interests include computer vision, computer-aided diagnosis, and pattern recognition. She has published approximately 15 academic papers through related journals and conferences. She also chairs several research projects, one of which is headed by the National Natural Science Foundation of China.

    Wei Xiong received his B.S. degree from the Department of Computer Science and Technology, Central China Normal University, Wuhan, China, in 2004, and his Ph.D. degree from the Institute of Pattern Recognition & AI, Huazhong University of Science and Technology, Wuhan, China, in 2011. He is currently an associate professor with the College of Computer Science, South-Central University for Nationalities, Wuhan, China. At present, he is a visiting scholar at Deakin University, Melbourne, Australia. His current research interests include computer vision, network security, and pattern recognition. He has published approximately 10 academic papers through journals and conferences.

    Yin Zhang is an Assistant Professor of the School of Information and Safety Engineering, Zhongnan University of Economics and Law (ZUEL), China. He is an IEEE Senior Member since 2016. He is an Excellent Young Scholar at ZUEL. He is Vice-chair of IEEE Computer Society Big Data STC. He was a Poster-Doctoral Fellow in the School of Computer Science and Technology at Huazhong University of Science and Technology, China. He serves as editor or associate editor for IEEE Access, IEEE Sensors Journal, etc. He is a Guest Editor for Mobile Networks and Applications, Sensors, Multimedia Tools and Applications, Journal of Medical Systems, New Review of Hypermedia and Multimedia, etc. He also served as Track Chair of IEEE CSCN 2017, TPC Co-Chair of CloudComp 2015 and TRIDENTCOM 2017, etc. He has published more than 80 prestigious conference and journal papers. His research interests include intelligent service computing, big data, social network, etc.

    Ning Pan is Lecturer in School of Biomedical Engineering at South-Central University for Nationalities. He worked as a Post-Doctoral Fellow in School of Computer Science and Technology at Huazhong University of Science and Technology (HUST) for two years. He received his M.S. in Computer Software and Theory from Chongqing University (CQU) and his Ph.D. in Computer Science from HUST, in 2007 and 2013 respectively. His research interests include medical image processing, data mining and statistical learning.

    Zhuo Yu received his bachelor’s degree in computer science and technology from Zhongnan University of Economics and Law, Wuhan, China, in 2014. He is currently pursuing a master’s degree in computer science and technology at Zhongnan University of Economics and Law. From 2014 to 2015, he participated in research on image processing. His research interests are focused on image segmentation and image enhancement.

    Enmin Song is a professor of computer science at Huazhong University of Science and Technology, Wuhan, China, the director of Center for Biomedical Imaging and Bioinformatics (CBIB). He is also an expert of the Thousand Talents in Chinese National Plan. He has presided over the completion of the national high technology research program (863) goal oriented projects, international cooperation projects in science and technology, the National Natural Science Foundation project. His research interests include medical image processing, intelligent medical equipment, medical and health large data analysis etc. He has more than 100 academic papers published in international academic journals.

    Chih-Cheng Hung is a professor of computer science at Kennesaw State University (KSU), GA, USA. He was with Intergraph Corporation from 1990 to 1993 for research and development of remote sensing image processing software. He is also with the Henan Key Laboratory of Oracle Bone Inscriptions Information Processing in Anyang Normal University, China, and the Laboratory for Machine Vision and Security Research in KSU. His research interests include image processing, pattern recognition and artificial intelligence.

    View full text