You are on page 1of 5

ET4393

Dali ZHANG

4219058

Review of Fast Accurate MR Thermometry Using Phase Referenced Asymmetric Spin-Echo EPI at High Field

ET4393

Dali ZHANG 4219058

2013/7/7

ET4393

Dali ZHANG

4219058

Abstract
In the paper, the authors applied frequency-selective gradient reversal spin-echo (SE) to referenced asymmetric spin-echo (ASE) MR thermometry. Relevant simulation, phantom heating experiment, and in vivo experiment were presented in the paper. It was proved that the technique was of high speed, high robustness, excellent chemical selectivity, and low sensitivity to physiologically related fluctuations. However, results also showed a slight underestimation caused by MR phase variation.

Introduction
Proton resonance frequency (PRF) of water changes with variation in temperature, which provides MRI with the capability of generating accurate temperature maps. However, any changes of local frequency will cause error in temperature measurement. So results of MRI thermometry should be corrected. Authors of the paper employed a correction approach with reference. The reference was chosen with a PRF-temperature dependence that was different from waters. After measuring temperature of water, they measured temperature of reference material at the same location. These two results had the same error components, so error in their difference would be rather small, as long as delay between these two measurements had been short. In order to shorten the delay, the authors implanted the imaging system with gradient reversal ASE technique and echo planar imaging (EPI). To prove that the technique worked properly, the authors first investigated frequencyselectivity and phase shifting effect of the system with Bloch simulation. Afterwards, phantom heating experiment took place. And finally, nice results of in vivo temperature stability measurement proved that the technique was a successful step in MRI thermometry. This review will first summarize the paper, including its experiments, results, and discussion. And based on the paper, a possible future step will be provided at the end of this review.

Methods
As shown in Fig.1, MR raw data was acquired from standard Siemens SE EPI sequence on a 7 T whole body scanner. EPI readout was shift in time in order to acquire phase sensitivity. Raw data was then transferred to ODIN software environment. Both magnitude and phase signals of water and reference were transferred into time-series with the same form. And after correction happened afterwards, water temperature maps would be generated.

Fig.1. Grad ient Reversal ASE Technique

ET4393

Dali ZHANG

4219058

For phantom heating experiment, a spherical phantom with 1% agar, 1% NaCl , 34% dimethyl sulfoxide (DMSO), and 64% water was used as the measurement target. Compared with water-fat mixture, this phantom had a less frequency shift, which leaded to strong constraint requirements. The authors made effective compensation for this by properly arranging miscibility of water & DMSO and setting long T2 and T2*. In the experiment, the ASE gradient reversal technique shown in Fig.1 was used for both imaging and heating. Relevant parameters are as follows: TE=21 ms, TEeff=19 ms, TR=4 s, BW=752 Hz/Px, resolution 128128, nominal voxel size 1.51.51.5 mm3 , GRAPPA factor 4, and total acquisition time 133 min. For in Vivo temperature stability measurement, a healthy brain was under test. Fat was regarded as the reference. However, with fat and water at the same position under rapid data acquisition, fat signal might not be recovered when data was acquired, which would cause a low signal-to-noise ratio (SNR) of fat. To achieve a high SNR in both the water and the reference images, no overlap between water image slice and either the excitation or refocusing slice of the unwanted material was allowed. The authors successfully managed to do this by adding RF pulse duration, which could help increasing separation between excited or refocusing slice of unwanted material. The RF pulse duration chosen by them were 4 ms, and 6.4 ms for higher robustness to B0 . Relevant parameters are as follows: TR 250~2000 ms, TESE = 21 ms, k-space centered at 33.6 ms, TEeff=12.6 ms, BW=1502 Hz/Px For temperature change calculation, C++ ODIN libraries were employed. They combined magnitude and phase images into complex images, and split up interleaved water and reference time-series. The authors assumed that only TEeff was affected by B0 and temperature and only TEeff changes with the Larmor frequency. Since location of channels will cause variance in receive phase, the authors divided the channel-specific reference data by the data from the complex image time-series of each of the eight individual receive channels, both water and reference data separately. Single time-series were then formed with magnitude-squared weighting. Afterwards, uncorrected temperature map was generated according to the PRF phase shift method using the water phase time-series maps. For temperature correction with a reference substance, it was known that Larmor frequency of the reference was different from waters, so image should be spatially shifted. In phantom experiment, DMSO was mixed with water, so reference-substance time-series had noise. The authors smoothed it with Gaussian kernel of 4 voexls. For in vivo experiment, fat does not overlap with water, time-series were fitted by a two-dimensional linear function for each time-step and slice. Afterwards, phase changes of the references were transferred into temperature maps the same form as waters, the correction was performed by subtracting references maps from waters temperature maps. Other important issues are Bloch simulation and B0 and B1 measurements. Bloch simulation was employed to investigate frequency-selectivity and phase shifting effect. To measure B0 and ensure that the achieved flip angle for the excitation pulse was no greater than 120 , modified actual flip angle technique was used.

ET4393

Dali ZHANG

4219058

Results
Simulation and experiment results of Bloch simulation, phantom heating experiment, and in Vivo temperature stability measurement were presented in the paper. According to these results, the technique implemented in the paper performed well, though with some slight errors. For Bloch simulation, signal magnitude had max at flip angle of 90, and dropped at higher and lower flip angle. With RF pulses duration of 6.4 ms, frequency bandwidth was only about 300 Hz, net transverse magnetization outside the imaging bandwidth of well below 1% of the achieved maximum. And off-resonant species was below 0.1% of the maximal achieved signal for the on-resonant species. In the simulation, phase of MR should be constant. However, results showed that it was changing with frequency offset and flip angle, which finally leaded to a slight underestimation of temperature change. For phantom heating experiment, two sensors were put into the phantom, one in the center, the other at the edge. For uncorrected results, there was a steady drift to higher result. After correction, the MR thermometry results matched the sensors results well with an underestimation of 0.5%. For in Vivo temperature stability measurement, correction method showed a good result with TR=2 s. When TR dropped, a decrease of SNR by a factor of 3 for water measurement was observed. For fat, however, due to shorter T1 , SNR hardly changed with TR. Results also showed that for the location where there were more fat, the result would be better. And for the brain under test in the experiment, correction worked well at inferior part, and did not work well at superior part.

Discussion
According to the authors, advantages of the technique implemented in the paper were as follow. Firstly, the correction technique provided low error temperature measurement results, only 2.2% with 100 flip angle and 6.4 ms RF pulses duration, and 0.9% with 4 ms RF pulses duration. Secondly, the technique successfully shifted most fat signal from water signal during measurement, which yielded an excellent chemical selectivity. Thirdly, the technique shortened the delay between water and reference test, which provided the correction method a better result and also reduced its sensitivity to physiologically related fluctuations by increasing detection speed. Last but not least, the technique also showed nice robustness, especially low sensitivity to B1 inhomogeneties. On the other hand, drawback of the technique was that the MR phase of SE varies with flip angle and frequency offset. It would lead to a systematic error in the measured temperature change. And the error finally turned out to be an underestimation of temperature change. To eliminate the error completely, the authors suggested that different TEeffs should be used, or the sequence-induced phased shift should be corrected with the SE phase shift simulation

ET4393

Dali ZHANG

4219058

depending on B0 and B1 at each point.

Possible Improvement
As it was proved in the paper, shortened delay between water temperature measurement and reference temperature measurement would result in lower error. To further shorten the delay, or even measure temperature of both water and reference at the same time, structure in Fig.2 might be a possible choice. As it is shown in Fig. 2, the system has 2 detectors Det1 and Det2, region of interest (ROI) is always put at the center between Det1 and Det2. And Det1 and Det2 should be installed on a line that is parallel to B0 . Water and reference will be excited at the same time, and Det1 would detect PRF of water, while at the same time, Det2 would detect PRF of the reference. In this way, the delay would be minimized, and the reference would not suffer from the SNR problem described in the paper. For further improvement, one may use Det1 for reference and Det2 for water soon after the measurement described above, and use the average or differential result. However, this might lead to problem similar to the reference SNR problem in the paper.

Det1

B0

ROI1

ROI2

ROI3

Det2
Fig.2 Possible Improvement

Conclusion
The paper demonstrated that ASE imaging using a simple frequency selection method can perform highly accurate referenced MR thermometry in phantoms and in vivo at 7T. The approach provided a successful step with in vivo SAR monitoring, and may lead to further development in MRI application.

You might also like