Phase contrast proton resonance frequency shift (PRFS) quantification based on Gradient Echo techniques is the prevailing standard when it comes to MR based temperature monitoring during MR-guided thermal treatments¹. In order to quantify temperature changes in adipose tissue, alternative MR properties, such as T₁ or T₂ relaxation, need to be measured. The presence of a heat applicator like the transducer of a High-Intensity Focused Ultrasound (HIFU) setup, or a radio-frequency (RF) applicator consisting of an array of RF antennas inside the MR scanner usually entails the usage of the body coil, resulting in rather low signal to noise ratio (SNR) values. Balanced steady-state free precession (bSSFP) offers larger SNR values than a Spoiled Gradient Recalled Echo (SPGR) readout. Moreover, inversion recovery (IR) or saturation recovery (SR) prepared bSSFP is widely used for fast T₁ mapping^2,3,4. In a previous work it has been shown that by choosing an asymmetric relation of echo time (TE) and repetition time (TR) the signal phase becomes sensitive to changes in resonance frequency⁵. In this work the feasibility and challenges of simultaneous PRFS and T1 mapping using bSSFP are demonstrated with a cooling down experiment of a phantom containing agar and a piece of pork lard.

The phantom consisted of a piece of pork lard which was put into a glass plunger. A liquid solution of agar (3%) was added to it and solidified thereafter. Two wooden skewers were used for guiding the fluoroptic temperature probes (Luxtron FOT Lab Kit, LumaSense Inc, Santa Clara, CA) (Fig. 1c)).

To investigate both T1 mapping quality and expected phase behavior, Bloch simulations were performed using MATLAB (MathWorks, Natick, MA, USA) (Fig. 2). To retain sufficient effect of the inversion preparation TR should be no longer than 7ms. The shortest possible TE for full k-space sampling was 1.9ms. The frequency shift of 0.009 ppm/°C translates into a shift of 1.2775 Hz/°C at 3 Tesla⁶. This results in a temperature sensitivity of approximately 0.01 rad/°C in the range of linear phase evolution around the resonance frequency, as shown in the simulation (Fig. 2). MR Imaging was conducted on a GE 3T 750w Scanner (Waukesha, WI, USA). A 12-channel GEM head coil was used during the cooling down experiment. The phantom was placed into a hot water bath before scanning in order to heat up. The phantom was subsequently placed into the scanner such that an axial slice covers both temperature sensors (Fig.1). During cooling down 2D T1 mapping with selective inversion on fat frequency followed by bSSFP readout was done. TE was chosen to be the minimum value for full Fourier sampling, which was 1.9ms for this setting (image resolution 128x96, FOV = 22cm, slice thickness = 4mm, flip angle = 35°). Inversion times were chosen to be 260ms, 310ms, 410ms, 610ms, 860ms and 1210ms. T1 maps were generated using a trust-region curve fitting algorithm implemented in MATLAB (MathWorks, Natick, MA, USA) fitting to the signal equation $$$M(TI_k) = A –B *exp(\frac{-TI_k}{T_1})$$$.

Phase difference calculations were done using the image at the earliest measurement time point, at the hottest temperature, as the reference image. Out of the 6 inversion recovery images the last one (TI=1210ms) was chosen respectively. The phase evolution in a ROI close to the temperature probe (red) does not correspond to the simulated values (blue) for this temperature range (Fig. 3b)). However, it is expected that the well-known B0 drift overlays the temperature induced phase shift. As a correction step, the phase of the pork lard (orange curve in Fig. 3b)) in a small ROI which is close to the agar ROI is traced and subtracted from the phase measured in the agar ROI. The corrected phase curve is now very close to the simulated one (violet curve).

The increase of T₁ with temperature correspond to the curve found at 1.5 T⁷. The T¹ values are higher, as expected at 3 T. The mean value inside the ROI (Fig. 4a)) depending on the measured temperature is plotted in Fig. 4b).

A rather homogeneous B0 field within the image is assumed in order to be located in the step-by-step linear phase curve. The presence of heat applicators, like ultrasound transducer or RF antennas, introduces B0 inhomogeneity, which will introduce banding artifacts. Phase cycled bSSFP acquisition is a solution in order to overcome these artifacts. In this way, the non-linear phase regions will be covered by a linear curve in the second image.IR prepared bSSFP is able to resolve PRFS phenomenon, thus enabling temperature quantification in both adipose and aqueous tissues at the same time.