Phase drift correction methods in MR temperature mapping are based on computing correction maps by polynomial fitting of phase shifts at reference points. Such reference phase shifts are measured either from extra oil phantoms placed around the imaging object or from fatty tissues inside the object. However, the use of oil phantoms necessitates a large FOV thereby limiting the spatial resolution of MR images, and it could limit oblique scanning. Correction with phase reference at fatty tissues has also limitations in low fat region apart from need of a special pulse sequence to separate fat region from water region.

In this work, we explore the use of magnetic field monitoring (MFM) to correct main field drift effects even at the region of low fat concentration, avoiding the use of extra phantoms and potentially allowing oblique scanning.

Temperature measurement in MR thermometry is based on the proton resonance frequency (PRF)¹ shift of water molecules, which states that the temperature change ΔT in an object is proportional to the phase difference Δϕ from the reference time. The temperature change is given as:

$$$ \triangle T=\frac{\triangle\phi}{\alpha TE \gamma B_{0}} $$$ (1)

where γ is the gyromagnetic ratio, B₀ is the main magnetic field strength, and α is the thermal coefficient of proton resonance frequency shift of water molecule (-0.01 ppm/°C). Phase drift is corrected by subtracting the correction map Δϕ_B from equation 1:

$$$ \triangle T=\frac{\triangle\phi -\triangle \phi_{B}}{\alpha TE \gamma B_{0}} $$$ (2)

The maps Δϕ_B were computed from the frequency shifts of FID signals, acquired by an array of MFM probes^2-3 attached around a Birdcage coil. For comparison proposes, we also placed four oil tubes around the object to get reference phases for computing the correction maps.

Figure 1 shows the MFM and MR thermometry pulse sequences. First, FID signals were acquired with four MFM probes, which were excited with a 1W rectangular pulse with a pulse length of 11us. The data acquisition of 10,000 samples was done with a sampling interval of 10us. Phase information was obtained from a gradient echo sequence with TR/TE of 150/10ms, flip angle of 60° and field of view of 230x230 mm². This sequence was repeated 8 times every 5 minutes. Images were acquired from a tissue-mimicking agar phantom⁴ that has electrical conductivity of 0.1 S/m at the background region. In the middle of the phantom was a tumor-tissue-mimicking insert that has electrical conductivity of 0.6 S/m. In addition, an optic fiber temperature sensor was inserted to the center of the phantom to evaluate accuracy of MR thermometry.

Phase correction maps were computed by bilinear interpolation of the frequency shifts at MFM probes or the phase shifts at the oil tubes, and they are shown in Fig. 2. The correction maps obtained with MFM are similar to the oil reference correction maps.

In the temperature maps without correction (Fig. 3A), the effect of phase drift is evident. The temperature change at the end of 40 minutes experiments is about 2°C even though the optic fiber temperature sensor reads constant temperature at the middle of the phantom. In addition to that, the temperature change is space variant. After applying Eq.2, the temperature map is corrected as shown in Fig. 3C whereas Fig. 3B shows the temperature maps corrected by the oil reference method.

Figure 4 compares the temperatures measured with the optic fiber sensor and the MR thermometry with and without correction. The temperature at the center of the phantom remained constant at 18°C during the 40 minutes when read by the optic fiber sensor. For an ROI of 3x3 located at the center of the phantom, the average temperature over time for the PRF method without correction was 17.6° with the standard deviation of 0.26°; whereas the average and standard deviation for the corrected maps with oil-reference method and MFM reference method were 18.03°± 0.24 and 17.95°± 0.07, respectively.

We introduced a method to correct the phase drift effects in MR thermometry by using magnetic field monitoring. The proposed method can be used in combination with any MR thermometry pulse sequence without making any changes in the sequence. In addition, the use of MFM probes allows to have better resolution because the reference points are not inside the FOV. This method has also a potential to make oblique temperature maps if more MFM probes are placed surrounding the desired FOV.