Rapid Magnetic Resonance Imaging of Blood Velocity and
Cardiac Function

Mark Doyle

Division of Cardiology, Allegheny General Hospital, Pittsburgh, PA, USA

Magnetic resonance imaging (MRI) has been applied to image the beating heart and is capable of dynamically imaging blood flow velocity. Currently, MRI technology has advance to the point where these data sets can be acquired in a routine manner during a breath-hold (i.e. 10-20 seconds). Further, advances are being widely pursued to allow real-time MR imaging. However, it is widely recognized that there exists a trade-off between spatial and temporal resolution, and achievement of rapid scant times is often achieved at the expense of reducing resolution below the optimum. This is particularly true for velocity imaging, where it is necessary to encode additional information. We have developed a rapid imaging approach applicable to dynamic MRI that allows reduction of the dynamic scan time, allowing either faster imaging, improved resolution or improved encoding of velocity data. The approach developed is termed BRISK (Block Regional Interpolation Scheme for k-space).

The BRISK approach reduces the acquisition time for cine sequences to about one third of the conventional time. The time reduction for BRISK has its origin in two features of dynamic cine MRI: 1) each frame of an MRI series is acquired directly in the Fourier domain (i.e. the MRI signal requires Fourier transformation to convert it into an image) which is referred to as “k-space” and 2) successive frames from a cardiac cine series typically have many features in common, i.e. a cine series has a dominant static component superimposed on a smaller dynamic component. By performing Fourier analysis of successive cine k-space data sets along the temporal direction, the nature of the static and dynamic components is revealed. It was noted that to a first approximation, successively higher dynamic frequency information tends to progressively be concentrated towards the center of k-space. BRISK accomplishes reduction of the scan time by applying a temporal distribution of data sampling based on this observation, such that towards the center of k-space the highest dynamic sampling rate is applied, and towards the periphery of k-space progressively lower temporal sampling rates are applied. Data that are not directly sampled are generated during post processing by performing temporal interpolation using Fourier methods. Thus, after processing the BRISK data, a complete k-space data set is produced for each time point in the cardiac cycle. For BRISK, each k-space data set contains both directly acquired and temporally interpolated data.

To accomplish rapid cardiac imaging, a number of innovations are simultaneously applied, including segmentation scanning (i.e. combining successive acquisitions of k-space into a single time point within the cardiac cycle), parallel imaging methods, and inherently fast scan techniques such as the steady state free precession approaches. BRISK is compatible with each of these approaches and can be additionally incorporated to further reduce scan time. BRISK has been applied in conjunction with segmented steady-state-free-precession cine imaging to allow three separate cine images to be acquired in the time generally required to obtain a single cine acquisition. Quantitative analysis were performed on the images to show that BRISK did not significantly alter the volumetric cardiac information.

While rapid imaging has potential to improve patient throughput and is a convenience for both patients and physicians, the fundamental data content has not been altered. However, an application area where BRISK has been applied to improve data content is Velocity Encoded Cine (VEC) imaging. When encoding velocity information, it is necessary to encode a reference image set and a velocity encoded data set for each velocity direction considered. Thus, the inherent amount of data to be encoded for velocity imaging is two to four times higher than conventional imaging. Further, the requirement to encode velocity data requires that inherently slower imaging sequences be applied, further reducing the speed with which data are acquired. Thus, the application of BRISK to velocity imaging allows multi-dimensional data to be acquired in a convenient breath-hold time. Without BRISK, either the acquisition time would be extended beyond that of a typical breath-hold (associated with reduced image quality) or data would be temporally compromised.

Examples of BRSIK applied to rapid, high-resolution conventional cine imaging and velocity imaging will be shown.