Time-resolved whole-heart cardiac imaging using highly parallel magnetic resonance

Lead Research Organisation: University College London
Department Name: Medical Physics and Biomedical Eng

Abstract

The broad aim of our research project is to develop new techniques to improve the quality of Magnetic Resonance Imaging (MRI). Our project specifically looks at this in relation to imaging the heart and surrounding area (cardiac imaging). MRI is a safe diagnostic tool providing good images of soft tissue organs and is used routinely for the imaging of static structures such as the brain. However it is not yet widely used for cardiac imaging because MRI is very sensitive to motion. The movement of the beating heart reduces image quality producing blurring and ghosting. A similar effect occurs when someone moves while you are taking their photo, though the mathematics of how the image is affected is different. There are two independent sources of motion: the cardiac contraction (heart beating) and respiratory motion (breathing). It is possible to compensate for the motion associated with cardiac contraction, but respiration is less predictable and consequently harder to compensate. Most cardiac MRI studies are therefore done while the patient holds their breath. Naturally there are limits to how long patients can do this for and therefore the quality of the images can be compromised. A typical cardiac MRI exam requires multiple breath holds and it has to be planned very carefully to ensure that the necessary images are acquired. This planning and the subsequent image analysis require highly trained staff and these are not available at most sites. This is one of the main obstacles for the widespread use of cardiac MRI. In this proposal we aim to develop new, faster and easier ways of acquiring cardiac MR images. We aim to do this by replacing existing 2D methods with 3D techniques. The advantage of this approach is that the whole heart can be imaged during a single acquisition and very little planning is required. To achieve this it is necessary to overcome some of the problems associated with respiratory motion so that the images can be acquired without the patient having to hold their breath. During the acquisition process the motion of the heart due to respiration will be measured and these motion measurements will then be taken into account when forming the images. We will take advantage of the fact that new MRI scanners can now acquire multiple chunks of data (using multiple receive devices) at the same time ( highly parallel imaging ). This enables acceleration of the acquisition and it can provide complementary information about motion. This new technology will complement and facilitate our research in several ways. For example, we will be receiving more information about the effects of the motion on the image thus allowing us to better correct the images. In certain situations, there is a limited amount of overall scan time available for the MR examination. This includes acquisitions where a contrast agent (dye) is injected into the patient. In order to make the best use of the time available for this type of scan we want to avoid acquisition of redundant information. To accomplish this it is necessary to rely on prior information about the movement of the heart and this can potentially degrade the reconstructed images. In this proposal we wish to investigate how we can rely less on prior information, this will involve complicated mathematics to determine the optimal balance between prior information and parallel imaging principles.In summary, the purpose of this proposal is to develop new MR imaging strategies that allow 3D imaging of the heart during free-breathing and under time-constraints. The developed imaging sequences will be tested on both volunteers and patients.

Publications

10 25 50

publication icon
Hansen M S (2008) Interactive Adjustment of Regularization in SENSE and k-t SENSE using Commodity Graphics Hardware in Interactive Adjustment of Regularization in SENSE and k-t SENSE using Commodity Graphics Hardware

publication icon
Hansen MS (2008) Cartesian SENSE and k-t SENSE reconstruction using commodity graphics hardware. in Magnetic resonance in medicine

publication icon
Kowalik GT (2012) Real-time flow with fast GPU reconstruction for continuous assessment of cardiac output. in Journal of magnetic resonance imaging : JMRI

publication icon
Odille F (2010) Model-based reconstruction for cardiac cine MRI without ECG or breath holding. in Magnetic resonance in medicine

publication icon
Odille F (2010) A parallel computing framework for motion-compensated reconstruction based on the motion point-spread function in A parallel computing framework for motion-compensated reconstruction based on the motion point-spread function

publication icon
Odille F (2010) On the accuracy of nonrigid motion correction in Workshop on Motion Correction,Kitzbuhel, Austria.

publication icon
Odille F (2009) Cardiac and respiratory motion compensated reconstruction driven only by 1D navigators in Cardiac and respiratory motion compensated reconstruction driven only by 1D navigators

 
Description The broad aim of our research project was to develop new techniques to improve the quality of Magnetic Resonance Imaging (MRI). Our project looked at this in relation to imaging the heart and other organs affected by respiratory motion. MRI is a safe diagnostic tool providing good images of soft tissue organs and is used routinely for the imaging of static structures such as the brain. It's use for cardiac imaging has been limited because MRI is very sensitive to motion. Uncorrected, the movement of the beating heart reduces image quality producing blurring and ghosting. A similar effect occurs when someone moves while you are taking their photo, though the mathematics of how the image is affected is different. There are two independent sources of motion: the cardiac contraction (heart beating) and respiratory motion (breathing). In some situations it is possible to avoid the motion associated with cardiac contractions by gating to an ECG, but respiration is less predictable and consequently harder to compensate. Most cardiac MRI studies are therefore done while the patient holds their breath. Naturally there are limits to how long patients can do this for and therefore the quality of the images can be compromised. A typical cardiac MRI exam requires multiple breath holds and it has to be planned very carefully to ensure that the necessary images are acquired. This planning and the subsequent image analysis require highly trained staff and these are not available at most sites. This is one of the main obstacles for the widespread use of cardiac MRI.



This joint grant was managed by bi-weekly team meetings that proved highly effective in developing new ideas, facilitating inter-disciplinary links and fostering translation to clinical systems. During the course of this grant we developed new techniques for acquiring the data and took advantage of the new 32 channel receive coil technology. This highly parallel data acquisition led to new image reconstruction methods that could infer motion from the data leading to quality images. Although the amount of data and computations increased, we were able to exploit the new opportunities provided by the availability of cheap, powerful graphics cards for the rapid computation of images. As measured by scientific output, the grant was highly successful with a total of 19 publications resulting from this work, of which 7 were in international refereed journals.
Exploitation Route The demonstrated MR imaging of the heart during free breathing and the real-time imaging of speech have potential clinical applications. Two of the papers from this grant make a significant contribution to the recent Open Source Gadgetron software for streaming image reconstruction (gadgetron.sourceforge.net). This software had over 1200 downloads in its first 15 months with a large international uptake, for example, 42% of these downloads were from the USA. The software has been interfaced to Siemens clinical MRI scanners for cardiac and interventional research applications.
[update March 2017: The Gadgetron continues to grow and be supported at NIH. The MR manufacturer Siemens now advertises that their product interfaces with Gadgetron. In addition a paper on a new vendor-neutral format for MR raw data (ISMRMRD) has been published with authors including Atkinson and Hansen].


The key contributions from this grant are twofold. Firstly, we introduced a new concept for interacting in real-time with algorithm parameters in magnetic resonance image reconstruction. Users can move a slider to interact with parameters previously regarded as "fixed". This enables them to determine the optimum settings which can vary for different applications. This real-time capability came about through the careful selection of algorithms appropriate to the physics, and an implementation on GPUs. Secondly, we provided a novel GPU implementation and adaptation of algorithms for reconstruction of MR images from radially sampled (non-Cartesian) data. Although fast to acquire, non-Cartesian data was previously slow to reconstruct. We enabled real-time reconstruction and visualisation of radially undersampled cardiac MR data, and, moving buffer displays of speech.
Sectors Digital/Communication/Information Technologies (including Software),Healthcare

 
Description Our findings continue to be used primarily through methodology published in peer reviewed journals. The grant funded some of the work that formed the basis of the Gadgetron software that is actively being used for the streaming reconstruction of MRI data.
First Year Of Impact 2011
Sector Healthcare
Impact Types Policy & public services

 
Description King's College London 
Organisation King's College London
Country United Kingdom 
Sector Academic/University 
Start Year 2007
 
Description PHILIPS HEALTHCARE 
Organisation Philips Healthcare
Country Netherlands 
Sector Private 
PI Contribution Feedback of our research results to the collaborator. Feedback on the collaborator's clinical and research products.
Collaborator Contribution Scientific advice, access to clinical scientists with knowledge of MRI hardware and software. Provision of pulse programming software environment and works in progress packages to enable clinical research.
Impact This collaboration enables the research we do using Philips MR systems.
 
Title Gadgetron 
Description Work on this grant provided some of the background for the Gagdetron Open Source software initiated primarily by Michael Hansen and Thomas Sorensen. The software is released under NIH's Open Source license. The software provides an architecture for the streaming reconstruction of MRI data allowing large datasets to be rapidly reconstructed. 
Type Of Technology Software 
Year Produced 2011 
Open Source License? Yes  
Impact The software has gained an international reputation and Siemens now market their MRI scanners as compatible with the software. It has helped to define a vendor-neutral standard for MR data reconstruction. Development is active and ongoing with an international collaborative network meeting weekly via Google Hangouts. 
URL http://gadgetron.github.io