4D-Flow Phase Contrast Magnetic Resonance Imaging has been shown useful for fundamental research and clinical applications. This technique could derive useful insights about the cardiovascular function and function changes due to diseases in murine models. So far long scan times, low SNR, low resolution and image artefacts compromise image quality and therefore the ability to use this method. We present an improved steady state sequence design in combination with a cryogenic coil and a sophisticated reconstruction process. It allows to characterize aortic blood flow in comparable quality to humans and to derive common quantitative parameters.For the preclinical MR imaging a Bruker Biospec 70/20 USR equipped with a two channel transmit/receive cryogenic (26 K helium cooled, Bruker) mouse head surface coil was used. This coil shows a factor 3-5 improvement in SNR compared to conventional room temperature coils.¹ Physiological observation and sequence gating was done with a SA Instruments 1030 (Stony Brook, NY 11790) monitoring device. All animals (n = 3) were measured under the approval of the ethics committee (Ref: G-14-91). A balanced four point phase contrast sequence from the manufacturer was redesigned to keep the magnetization in a steady state, by an online trigger decision using a double gating strategy. The gating device was modified to generate two trigger outputs, one to determine the respiration state and the other for the R-peak of the ECG. The steady state is interrupted a few millisecond when detecting the next R-peak from the ECG. The data was reconstructed using Matlab. In order to be able to segment and display smaller vessels a POCS algorithm was implemented in the magnitude reconstruction. A multilevel wavelet based segmentation algorithm with a hard threshold was developed. It automatically segments the aorta in 3D in every cine frame from the magnitude data and detects static tissue that is used for eddy current correction. Results with Daubechies wavelets showed a good agreement with a manual segmentation. The phase offset in the velocity data was corrected by a 3D fitting algorithm based on Legendre Polynoms.² The data was then fed in our default flow processing pipeline used for human flow post-processing as described in literature.^3,4 The data was visualized in Ensight (CEI Inc). Parameters: FOV = 20x16x14 mm³, TE/TR = 2/5; Matrix = 60x50x50 (interpolated to 125x100x88), VENC = 150 cm/s; FA = 15°, Acquisition time in vivo ~ 45 min, POCS iterations = 20.

It was possible to acquire the blood flow in the aorta with a very high resolution showing complex flow patterns even in the aortic branches. The images show an uniform signal intensity over the cardiac cycle. The wavelet segmentation allowed a magnitude segmentation mask in every time frame and could differentiate static tissue from moving tissue. It was possible to remove the phase offsets by means of the 3D phase offset correction based on Legendre Polynomials. Common parameters to quantify the blood flow were derived. For example peak speed 127 ± 29 cm/s, flow through a plane in the ascending aorta was equal to 62 ± 15 ml/min.

With the new steady state sequence design it was possible to remove commonly neglected steady state artefacts and to achieve an uniform signal intensity over the cardiac cycle. This facilitates the segmentation process as it is based on hard thresholding. Higher SNR due to the cryogenic coil, redundantizes signal averaging. Thus measurement time is reduced by a factor of three compared to published data⁵ (at equal resolutions). The wavelet segmentation allowed a magnitude segmentation mask in every time frame and a superior static tissue detection compared to conventional 2D methods (based on standard deviation over time). Even the carotid arteries are well segmented which are not covered with thresholded speed-sum-squares masking methods from the velocity information. Several improvements in sequence design, image reconstruction, post-processing in combination with a highly sensitive coil enables to achieve a sufficient image quality for murine high resolution 4D-Flow. Acceleration methods like GRAPPA, SENSE could be used to speed up the acquisition. However the resulting loss of SNR could impede flow quantification in small vessels.

The presented framework - to our knowledge- allows for the first time to generate data quality high enough to investigate preclinical questions. Our framework -to our knowlege- allows for the first time to study complex flow patterns at high spatial resolutions. It will permit to study flow changes non-invasively in mouse models. Thus allowing for example to investigate the development of atherosclerotic plaques or valvular heart disease, which could provide a better understanding of these lethal developments.