Image analysis for whole-body MR imaging

Stichting Technische Wetenschappen (STW), grant 10894

Oleh Dzyubachyk, PhD
Jorik Blaas, PhD
Marius Staring, PhD
Rahil Shahzad, PhD

Introduction

The rapid progress in MR scanning technology increasingly enables imaging the whole body, and first clinical applications of whole-body MRI are rapidly emerging, mainly for cardiovascular risk assessment, and in oncology. For instance, vascular imaging protocols have been developed that enable a vascular checkup and risk assessment for cardiovascular disease. These do not only enable imaging the entire vascular system (including carotids, aorta, renal arteries and peripheral vasculature), but also the amount of body fat and its distribution over the body. Especially the presence of excessive fat in the abdomen is an important risk factor for the onset of vascular diseases, and in combination with the vasculature, whole-body MR may provide a good assessment of the risk for a patient to develop vascular disease. As such, these protocols have recently been added to large scale epidemiological imaging studies with thousands of participants. Also for oncological applications, whole-body MR protocols have been developed for detection of cancerous lesions, and for cancer stageing based on for instance bone marrow involvement in hematological cancers. These protocols have been shown to enable evaluation of treatment effect, shortly after the onset of chemotherapy regimen or radiation effect, and as such in the future may form a viable alternative for PET scanning.

As described above, the introduction of these new whole-body MR imaging techniques will provide important clinical benefits. However, there are several technical challenges inherent to whole-body imaging that so far only been sparsely addressed due to the relative novelty of whole-body MR acquisition. For instance, whole-body scanning is typically performed in single parts that should be stitched together to form a whole-body scan: this is often performed manually. Also, visual analysis of all the acquired image data is difficult due to the large amount of data in a whole-body MR examination. Accurate quantification of vascular abnormalities and fat distribution from MR data is important, but also the identification and labeling of anatomical structures should be automated. Moreover, rapid automated detection of abnormalities is required in order to direct the interpreting physician towards diseased areas, preferably while the patient is still in the scanner. Finally, the coarse resolution of whole-body scans may necessitate the acquisition of higher resolution targeted scans to zoom in on suspected areas of pathology: the radiologist requires a “Google Patient” analysis tool that executes the extraction of clinically relevant information disease status from thousands of images while the patient is still in the scanner.

Goals

The goal of this project is to develop quantitative image analysis and visualization algorithms for exploration of whole-body MR imaging studies, with a focus on cardiovascular risk assessment and oncological treatment evaluation.

Results

Within this project, several smaller projects were defined and executed:

  • Volume reconstruction for multispectral whole-body MR data
  • Inter-station intensity standardization for whole-body MR data 
  • Whole-spine volume reconstruction
  • Interactive super-resolution reconstruction of MRI mouse data
  • Super-resolution reconstruction of cardiac MR
  • Complete arterial tree extraction from whole-body MRA 
  • Comparative volume visualization
  • DREAM-based shimming for whole-body 2-point Dixon’s water/fat MR

All of these project have a common theme that they consider different aspects of reconstructing a single volume from multiple MR acquisitions.

Volume reconstruction for multispectral whole-body MR data

In this project we developed a novel all-in-one method for reconstruction of a whole body volume from multi-station multi-spectral MR data, that incorporates, in particular, both inter-station intensity calibration and bias correction [13]. The relation between the channels of the multi-spectral data set is preserved by performing joint processing on all the available data channels.



Inter-station intensity standardization for whole-body MR data

In this project we developed and validated a method for performing inter-station intensity standardization in multi-spectral whole-body MR data [8]. Different approaches for mapping the intensity of each acquired image stack into the reference intensity space were developed and validated. The registration strategies included: “direct” registration to the reference station, “progressive” registration to the neighbouring stations without and with using information from the overlap regions of the neighbouring stations.


Whole-spine volume reconstruction

In this work, an automated method for reconstruction of the complete spine from multistation 7T MR data was developed [1,9]. The method consists of a number of image processing steps, in particular intensity inhomogeneity correction and image registration for recovery of unknown interscan bed translations, which result in high-quality spine volume reconstructions.


Interactive super-resolution reconstruction of MRI mouse data

In this work we introduced a novel local approach to super-resolution reconstruction (SRR) that aims to overcome the computational problems and allow researchers to efficiently explore both global and local characteristics in whole-body small animal MRI [4,10–12]. The method integrates state-of-the-art image processing techniques from the areas of articulated atlas-based segmentation, planar reformation, and SRR. A proof-of-concept is provided with two case studies involving CT, BLI, and MRI data of bone and kidney tumors in a mouse model. We showed that local SRR-MRI is a computationally efficient complementary imaging modality for the precise characterization of tumor metastases, and that the method provides a feasible high-resolution alternative to conventional MRI.


Super-resolution reconstruction of cardiac MR

In this project we developed and validated a method for improving image resolution of late gadolinium-enhanced (LGE) magnetic resonance imaging (MRI) for accurate assessment of myocardial scar [3,5,14]. A super-resolution reconstruction (SRR) technique was applied to the three anisotropic views: short-axis (SA), two-chamber, and four-chamber, to reconstruct a single isotropic volume. For compensation of the interscan heart motion, a joint localized gradient-correlation-based scheme was developed.
     



Complete arterial tree extraction from whole-body MRA

In this project we developed a fully automated algorithm for extraction of the 3D arterial tree and labelling the tree segments from whole-body magnetic resonance angiography (WB-MRA) sequences [7]. The algorithm developed consists of two core parts (i) 3D volume reconstruction from different stations with simultaneous correction of different types of intensity inhomogeneity, and (ii) Extraction of the arterial tree and subsequent labelling of the pruned extracted tree.



Comparative volume visualization

In this project we developed and validated a novel approach to the comparative visual analysis of whole-body MRI follow-up data [2]. Our method is based on interactive derivation of locally rigid transforms from a pre-computed whole-body deformable registration. Using this approach, baseline and follow-up slices can be interactively matched with a single mouse click in the anatomical region of interest. In addition to the synchronized side-by-side baseline and matched follow-up slices, we have integrated several techniques to further facilitate the visual comparison of the two datasets.



DREAM-based shimming for whole-body 2-point Dixon’s water/fat MR

In this work we showed the effect, efficiency, and image quality improvements achievable by Dual Refocusing Echo Acquisition Mode (DREAM)-based B1 shimming in whole-body magnetic resonance imaging (MRI) at 3T using the example of water/fat imaging [6,15].


Status

Finished
 

Publications

         Journal papers:
  1. O. Dzyubachyk, B. P. F. Lelieveldt, J. Blaas, M. Reijnierse, A. Webb, R. J. van der Geest, An Automated Algorithm for Reconstruction of the Complete Spine from 7 Tesla MR, Magnetic Resonance in Medicine 69:1777–1786 (2013).
  2. O. Dzyubachyk, J. Blaas, Ch. P. Botha, M. Staring, M. Reijnierse, J. L. Bloem, J. H. C. Reiber, R. J. van der Geest, B. P. F. Lelieveldt, Comparative Exploration of Whole-Body MR through Locally Rigid Transforms, Journal for Computer Assisted Radiology and Surgery, Jul;8(4):635 47 (2013).
  3. Q. Tao, O. Dzyubachyk, D. H. J. Poot; H. J. Lamb, K. Zeppenfeld, B. P. F. Lelieveldt, R. J. van der Geest. Journal of Cardiovascular Magnetic Resonance, 16(Suppl 1):O40 (16 January 2014).
  4. A. Khmelinskii, E. Plenge, P. Kok, O. Dzyubachyk, T. J. A. Snoeks, D. H. J. Poot, C. W. G. M. Löwik, Ch. P. Botha, W. J. Niessen, E. Meijering, L. van der Weerd, B. P. F. Lelieveldt, Interactive Local Super-Resolution Reconstruction of Whole-Body MRI Mouse Data: A Pilot Study with Applications to Bone and Kidney Metastases, PLOS ONE, 9 (9):e108730 (29 September 2014).
  5. O. Dzyubachyk, Q. Tao, D. H. J. Poot, H. J. Lamb, K. Zeppenfeld, B. P. F. Lelieveldt, R. J. van der Geest, Super-Resolution Reconstruction of Late Gadolinium Enhanced MRI for Improved Myocardial Scar Assessment, Journal of Magnetic Resonance Imaging 42(1): 160‒167 (July 2015).
  6. M. Hooijmans, O. Dzyubachyk, K. Nehrke, P. Koken, M. Versluis, H. Kan, P. Börnert. Fast multi-station water-fat imaging at 3T using DREAM-based RF shimming, Journal of Magnetic Resonance Imaging, 42(1): 217‒223 (July 2015).
  7. R. Shahzad, O. Dzyubachyk, M. Staring, J. Kullberg, L. Johansson, H. Ahlström, B. P. F. Lelieveldt, R. J. van der Geest. Automated Extraction and Labelling of the Complete Arterial Tree from Whole-Body MRA Data, Medical Image Analysis 24(1): 28‒40 (2015).
  8. O. Dzyubachyk, B. P. F. Lelieveldt, M. Staring, M. Reijnierse, R. J. van der Geest, Inter-Station Intensity Standardization for Whole-Body MR Data, Magnetic Resonance in Medicine, in press.

    Conference papers & Abstracts:
  9. O. Dzyubachyk, B. P. F. Lelieveldt, J. Blaas, M. Reijnierse, A. Webb, R. J. van der Geest, An Automated Intensity Correction, Registration and Volume Stitching Algorithm for Reconstruction of the Whole Spine from 7 Tesla MR Data, Proceedings of International Society for Magnetic Resonance in Medicine (ISMRM 20th Annual Meeting), Melbourne, Australia, May 5–11, 2012, pp. 750 (2012).
  10. A. Khmelinskii, E. Plenge, P. Kok, D. H. J. Poot, O. Dzyubachyk, Ch. P. Botha, E. Suidgeest, W. J. Niessen, L. Van der Weerd, E. Meijering, B. P. F. Lelieveldt. Towards Interactive Super-Resolution Reconstruction of Whole-Body MRI Mouse Data, Proceedings of International Society for Magnetic Resonance in Medicine (ISMRM 20th Annual Meeting), Melbourne, Australia, May 5–11, 2012, pp. 4285 (2012).
  11. A. Khmelinskii, E. Plenge, P. Kok, O. Dzyubachyk, D. H. J. Poot, E. Suidgeest, Ch. P. Botha, W. J. Niessen, L. Van der Weerd, E. Meijering, B. P. F. Lelieveldt. Super-Resolution Reconstruction of Whole-Body MRI Mouse Data: an Interactive Approach, IEEE International Symposium on Biomedical Imaging: From Nano to Macro (ISBI 2012), Barcelona, Spain, May 2–5, 2012. Proceedings edited by A. F. Frangi, A. Santos, R. Ober, D. Rueckert. IEEE, May 2012, pp. 1723 1726.
  12. A. Khmelinskii, E. Plenge, P. Kok, O. Dzyubachyk, D. H. J. Poot, E. Suidgeest, Ch. P. Botha, W. J. Niessen, L. van der Weerd, E. Meijering, B. P. F. Lelieveldt. Super-Resolution-Reconstruction of Whole-Body MRI Mouse Data. TOPIM-2012 (annual Winter Conference of the European Society for Molecular Imaging).
  13. O. Dzyubachyk, R. J. van der Geest, M. Staring, P. Börnert, M. Reijnierse, J. L. Bloem, B. P. F. Lelieveldt, Joint Intensity Inhomogeneity Correction for Whole-Body MR Data, Medical Image Computing and Computer-Assisted Intervention, Lecture Notes in Computer Science, vol. 8149, pp. 106 113 (September 2013).
  14. O. Dzyubachyk, Q. Tao, D. H. J. Poot, H. J. Lamb, K. Zeppenfeld, B. P. F. Lelieveldt, R. J. van der Geest, Improved Myocardial Scar Characterization by Super-Resolution Reconstruction in Late Gadolinium Enhanced MRI, Medical Image Computing and Computer-Assisted Intervention, Lecture Notes in Computer Science, vol. 8149, pp. 147 154 (September 2013).
  15. M. Hooijmans, O. Dzyubachyk, K. Nehrke, P. Koken, H. Kan, M. Versluis, P. Börnert. Improved fast multi-station water-fat imaging at 3T, Proceedings of International Society for Magnetic Resonance in Medicine (ISMRM 22th Annual Meeting), Milan, Italy, May 12–16, 2014, pp. 559 (2014).

Contact

Rob J. van der Geest, PhD
Division of Image Processing
Department of Radiology, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden, The Netherlands
Tel. +31 (0)71 526 2138
e-mail: R.J.van_der_Geest@lumc.nl