Synopsis
Sodium T1
maps from three different Variable Flip Angle (VFA) methods were assessed
against Inversion Recovery (IR) data, showing that the estimated method of
slopes (eMOS) is superior to both linear and non-linear fitting methods in the
healthy brain. eMOS requires two flip angle data sets to produce volumetric T1
maps of the brain, in comparison to other techniques which require 4 or more data
sets to produce maps, and therefore allows for shortening of the overall time
spent scanning. Furthermore, T1 maps may also give further insight
in to the underlying tissue structure surrounding sodium nuclei.Purpose
To develop
and optimize a B1-corrected variable flip angle T
1 mapping
acquisition strategy at 3T, producing volumetric sodium T
1 maps of
the brain.
Methods
6 healthy
male volunteers, mean age 30 years ± 5 years were scanned using a GE MR750
3T system (Waukesha, WI) with a birdcage 23Na/1H head coil (Rapid Biomedical, Rimpar,
Germany).
The
variable flip angle (VFA) acquisition was performed with a spoiled gradient
echo acquisition utilising a 3D cones readout [1]. FOV = 24cm, TE = 0.5ms, TR =
110ms, flip angles chosen from simulation of the sodium signal = 55, 75, 95, 105,
115 degrees, readout duration = 30ms, averages = 6, readouts per average = 197,
nominal voxel size = 4.5x4.5x4.5 mm3, scan time = 2
minutes.
The sodium
B1 acquisition scheme was performed with the same scan parameters as
the VFA sequence, however TR = 220ms, data acquired with flip angles = 50 and
100 degrees.
The IR-gradient
echo imaging was performed with an adiabatic inversion pulse, and a 3D cones
readout. FOV = 24cm, TE = 0.5ms, TR= 180ms, inversion times chosen from
simulation of the sodium signal = 19, 25, 30, 35, 40ms, readout duration = 30ms,
averages = 6, readouts/average= 197, nominal voxel size = 4.5x4.5x4.5mm3,
Scan time = 4 minutes.
A power
noise correction was performed on all of the sodium images 2. B1 maps were
constructed from the B1 dual flip angle data 3. The raw sodium images were
registered with a T1 weighted proton scan (TE = 3.18 ms, TR = 8.16
ms, 1.5 mm isotropic resolution (reconstructed to 1mm isotropic) in SPM8, and
the proton scan used to produce white matter, grey matter, and CSF masks, with masks
gated on p > 0.95.
VFA T1 maps were constructed with linear
fitting, iterative non-linear fitting, and eMOS, with and without B1 correction 4,5. The IR maps were constructed with non-linear fitting. Processing was
performed using Matlab.
Average sodium T1 values for grey matter,
white matter, and CSF obtained from the segmentation masks and T1 maps, over
the whole brain.
Relative
error maps between the VFA T1 maps and the IR T1 maps
were produced according to equation 1, and the average error for each tissue
type with and without B1 mapping was calculated, as seen in figure 1.
It is noted that, in comparison to Chavez, no site specific correction,
normally obtained from the difference between an IR T1 map and eMOS
T1 map, was applied to the eMOS T1 maps produced in this
study.
Equation 1
$$Error = 100\times(\frac{VFA_{T1Value}-IR_{T1Value}}{IR_{T1Value}}) $$
Results and Discussion
A two-tailed t-test was used to compare the difference in mean whole brain tissue T1 values, data shown in figure 2, between data sets corrected and uncorrected for B1 non-uniformity. The results show that there is no significant difference between T1 maps corrected or uncorrected for B1 at sodium frequency at 3T, with a birdcage head coil (p > 0.05).
A two-tailed t-test was used to compare the results from each individual VFA method with both Inversion recovery T1 values at 3T, shown in figure 2, and average relative error for grey matter and white matter T1 maps, shown in figure 1. The results of this test are shown in figure 4, showing that the eMOS method outperforms a linear fitting method for estimating T1 values in both grey matter and white matter (p < 0.05 for both). However, it can be seen that the non-linear fitting performs as well as eMOS for estimating white matter T1 (p > 0.05) but that eMOS performs better than non-linear fitting for grey matter T1 estimation (p < 0.05). Finally, it has been shown that there is no significant difference between linear or non-linear fitting for estimating T1 (p > 0.05 in both cases). Figure 3 shows a masked IR T1 map for each tissue type.
Conclusion
In conclusion, a new application of eMOS has been demonstrated in estimating accurate sodium T
1 maps, using only a pair of two-minute sodium scans, with eMOS outperforming other VFA methods in producing accurate T
1 maps. The study has also shown that the spatial variance of B1 is negligible in a birdcage head coil. T
1 mapping is useful to aid in optimisation of sequence TR and flip angle, as well as assessing tissue T
1 changes associated with change in microstructure. Further to this study a comparison of eMOS against progressive saturation techniques, assessing the method at higher spatial resolutions, and finally studying sodium T
1 in cases of pathology, will be undertaken.
Acknowledgements
We would like to thank the MRC for their funding.References
[1] Gurney
et al, Magnetic Resonance in Medicine, 2006, 3: 575-582.
[2]
Miller and Joseph, Magnetic Resonance Imaging, 1993, 11: 1051-1056.
[3]
Tofts, Quantitative MRI of the Brain, 2003, 2.7.1 B1 Field Mapping, 26-30.
[4]
Chavez et al, NMR in biomedicine, 2012, 9: 1043-1055.
[5] Deoni
et al, Magnetic Resonance in Medicine, 2005, 50: 237-241,