Scientists and clinicians interested in MR Thermometry.Radial k-space sampling is insensitive to motion due to its efficient coverage of k-space. However, radial sampling is particularly sensitive to trajectory errors, and thus is likely to cause image distortions. Recently, Trajectory Auto-Corrected Image Reconstruction (TrACR) ¹ has been proposed as a general method to reconstruct radial images free of trajectory errors. In this work, we first evaluated the effect of TrACR for radial temperature imaging. Then, we examined the feasibility of this method on dynamic temperature imaging with only one calibration in order to reduce the computation time based on the assumption that gradient errors are time-invariant ².

A cooling down process of a 1% agar phantom was scanned by an 8 channel head coil (Invivo Corp, Gainesville) after the phantom was heated to about 50℃. The data were acquired respectively by a normal radial sequence and a golden angle radial sequence on a Philips 3T system (Philips Healthcare, Best, the Netherland). An optic fiber was inserted into the phantom to collect the temperature data as a reference. Imaging parameters for the phantom experiment: acquisition matrix size = 80 × 80, FOV = 160mm×160mm, slice thickness=3mm, TR = 50ms, TE = 10ms, flip angle = 15°, number of projections=100, 160 k-space samples each projection.

TrACR is a method to correct trajectory errors based on SENSE parallel imaging reconstruction. Our evaluation and analysis were divided into two steps. In the first step, the fully sampled data were corrected by TrACR in which the calibration for trajectory correction was done in every time frame. In the second step, single-calibration TrACR was performed in which all images used the same trajectory estimation derived from the image of first frame. Temperature maps were calculated using proton resonance frequency shift method ³ based on a fully-sampled reference. Subsequently, the data without correction, and the data corrected by TrACR and single-calibration TrACR were evaluated and compared.

Temperature maps reconstructed from the initial data, the data corrected by TrACR and single-calibration TrACR with normal radial and golden angle radial are shown in Figure 1 respectively. Figure 2 shows and compares the temperature evolution curves. The curve corrected by TrACR and the curve corrected by single-calibration TrACR are closer to the reference than the curve without correction, especially with golden angle radial. Table 1 shows the temporal root-mean-square errors (RMSEs) of the temperature evolution curves showed in Figure 2. The difference between RMSEs of conventional TrACR and single-calibration is less than 0.05℃.According to our results, the trajectory errors can lead to obvious temperature errors, thus it is necessary to correct trajectory errors for accurate temperature imaging. Both of the TrACR and the single-calibration TrACR can correct the errors, and the results of golden angle radial sampling could be better corrected by these two methods. The single-calibration TrACR in this work can acquire similar results to the ones corrected by TrACR with calibration in images from every time frame, but costs less time due to fewer calibration processes. The results indicate that the single-calibration TrACR is feasible for dynamic temperature imaging and the assumption that gradient errors are time-invariant may be meaningful.