Improved excitation profile and image quality has been clearly demonstrated in our phantom and in-vivo studies by compensating for B0 and linear eddy currents separately. These methods are easy to implement. To collect eddy current correction data, theoretically only four acquisitions are needed on each axis. As shown in , the linear and B0 eddy current terms measured this way are fairly consistent with those calculated from data acquired with eleven gradients amplitudes and eight slice locations. The result can be further improved with the use of high performance scanners where calibration data can be obtained with higher SNR. Collecting data from more slices and gradient amplitude steps improve the robustness of the fitting process and give less noisy measurement. Acquiring eddy current correction data can be done within a few seconds since signal averaging that is often employed to improve the SNR of the measurements using methods presented in (12
) is no longer necessary. Since eddy current characteristics on the commercial scanners are usually relatively stable with time, the compensation data only needs to be updated periodically. The residual phases from the fitting process for the eddy current characterization show little correlation to the applied gradient as shown in . Therefore the assumption that higher order eddy current terms can be ignored here is reasonable. Possible sources contributing to residual phase linearly dependent on location may include long time constant linear eddy currents from the gradient preceding the test gradient and imperfect linear shimming. Residual phase independent of position may include long time constant B0 eddy current from the preceding gradient, or off-resonance effects.
The eddy current compensation mechanism for half pulse excitation works well when the imaging plane is orthogonal to one of the three physical axes, where eddy current compensated RF pulses and gradients can be pre-designed for imaging in either axial, saggittal or coronal plane. As both B0 and linear eddy currents scale with the gradient amplitude played on the three physical axes, the compensation can be easily adapted to excite a desired slice thickness on the scanner. Imaging an oblique slice may present a challenge for the proposed approach since redesigning of the RF pulse for oblique angle according to combination of the linear eddy currents on all three physical axes is required for each oblique angle. Correct timing between the RF pulse, RF phase modulation for B0 eddy current correction and the selective gradient is also crucial for obtaining good slice profile and is currently adjusted empirically. This can potentially be improved by adopting the delay measurement method presented in (19
Our phantom studies have demonstrated that B0 eddy current effects can also be corrected by removing the constant phase offset between the two acquisitions with opposite selective gradients polarities. The phase offset was extracted from the onedimensional slice profiles. Since the one-dimensional profiles can be obtained in as short as two repetitions, either approach can be used for B0 eddy current compensation.
To compensate for eddy current induced by readout gradient, the separation of B0 and linear eddy current terms using the proposed methods allows the use of linear operations to calculate trajectories at arbitrary angles from measurements made on the orthogonal axes when the trajectories only differ in position or rotation angle, therefore removes the need to collect eddy current correction data for each trajectory. In case B0 eddy currents can be ignored, the trajectories measured on the physical axes using the method in (13
) is then equivalent to k
) measured here and therefore can be linearly combined to calculate trajectories at other angles. Significant improvement in image quality has been demonstrated in (15
) with this approach. The measurement of linear eddy current effect with this method was also indicated in (14
). However, the effectiveness of these approaches highly depends on the B0 eddy current characteristic of the scanners used. This is shown in the phantom images acquired on the scanner used in this work, where the measured B0 eddy currents are about one order higher than that observed on other high performance scanners. shows images reconstructed with different correction strategies from the fourth echo of the same data acquired on the spherical phantom. Again without eddy current correction, the image in is significantly distorted. Improvement in image quality is seen in image reconstructed using k-space trajectories measured with method in ref (13
) () and image reconstructed using k-space trajectories measured with method in ref (13
) to account for linear eddy currents and B0 eddy currents measured with method in ref (14
) to account for B0 eddy currents (). However, image distortions are still seen in both images. The distortion-free image was obtained in with the approach in this work. Though the corrections shown here are for 2D radial readout, they can be easily adapted to other readout trajectories such as 3D projection reconstruction and spiral acquisition.
Figure 8 Phantom images reconstructed with different correction strategies. (a) Reconstructed without correction. (b) Reconstructed using k-space trajectories measured with method in ref (13). (c) Reconstructed using k-space trajectories measured with method in (more ...)
With effective compensation for both B0 and linear eddy current for half-pulse excitation and a multiple-echo readout, signal intensity images and R2* maps with improved quality have been obtained throughout the whole freezing process. However, subject motion during the in vivo experiments complicates the registration of the measured temperatures to the pixels in the collected 2D images, consequently relating the measured temperatures to the R2* values accurately to generate the calibration curve has been difficult. In addition, the motion observed in our images during cryoablation may also be an issue for accurately map R2*. Future work will optimize the experiment setup or pulse sequence to better visualize the thermal probes in the images for better localization of the temperature readouts and for motion compensation through image registration.
Our experiments were performed on a low filed strength scanner (0.5 T) with relatively low gradient performance. Therefore, relatively long RF pulse (1.6 ms) was used for excitation along with a readout bandwidth of 31.25 kHz. The readout time for each radial line was ~2 ms. Signal loss during the excitation and signal decay during readout for the short T2 components is therefore not negligible. T2 decay during the readout causes image blurring and consequently affects the achievable image resolution. Several methods have been proposed to improve the SNR efficiency of short T2 imaging (20
). Use of high field strength scanners that are friendly to intervention will significantly improved quality of the R2* mapping by providing higher SNR, higher temporal and spatial resolution, where shorter RF duration and readout duration can be used.
In conclusion, an effective approach to compensate for both B0 and linear eddy currents for half-RF pulse excitation is presented based on the measured actual B0 and linear eddy currents. The significant improvement in the excitation slice profile allows better characterization of short T2* tissues as shown in the cryoablation experiments. The proposed techniques should be useful in imaging tissues with short T2s in general, especially when quantification is desired. The eddy current measurement approach has also been applied to dramatically improve the image reconstruction by compensating for eddy currents induced by the readout gradient.