To investigate proton density (PD)- and T1-weighted zero TE (ZTE) imaging for temperature mapping in MR visible tissues such as cortical bone, muscle, or fat.

Several clinical thermotherapy techniques that raise the temperature of regions in the body for tissue ablation or mild hyperthermia gained traction in oncology. High intensity focused-ultrasound (HIFU) is used for treatment in brain disorders (e.g. glioblastoma, essential tremor, Parkinson). During transcranial HIFU under MR guidance, it is important to monitor temperature changes in the human skull as it absorbs ultrasound with ~90x greater efficiency than soft tissue¹. Cortical bone has ultra-short T₂ relaxation time (<1ms) and thus customized acquisition techniques are required to capture the rapidly decaying MR signals with sufficient signal to noise ratio (SNR)². Recently, ZTE MR bone imaging in the head was introduced³. Here, we propose to (i) extract relative temperature information from PD-weighted ZTE images and (ii) additionally calibrate the signal contamination from T₁ saturation effects.

According to Boltzmann distribution and Curie law, the magnetization M₀ depends on the thermal equilibrium, where the susceptibility is inversely proportional to the absolute temperature 1/T. For ZTE sequences with zero nominal echo time TE (TE<<T₂*), short repetition times (TR<<T₁) and small flip angles (α<<1), the steady state signal equation for gradient echo sequences can be approximated by $$$y=\frac{\alpha*M_0}{1+\beta}$$$ with $$$\beta=\frac{T_1}{TR}\frac{\alpha^2}{2}$$$ (Eq.1).

PD-weighting is achieved by using very small flip angles, whereas with increasing flip angles T₁ signal contamination gets more prominent. To investigate the achievable SNR based on the analytically predicted and optimized ZTE parameters, SNR efficiency was studied by acquiring 50 ZTE images in the human head using a single-channel Tx/Rx head coil at 3T (MR750w, GE, Waukesha, USA) (Figure 1a). The experiment was repeated 20 times using the body coil. A dedicated T₁/T₂ quantification phantom (DiagnosticSonar, Livingston, UK) consisting of 12 gadolinium-doped agarose samples representing a range of T₁/T₂ relaxation properties was used. The phantom was cooled to 6°C and placed inside the scanner overnight. Temperature imaging was performed as it warmed to room temperature by acquiring PD- and T₁-weighted ZTE images (Figure 1b). Ground-truth temperature was measured using fluor-optic probes (LumaSense Technologies, Santa Clara, CA, USA) positioned inside the setup. The overall change in temperature (ΔT) seen in experiments was ~9.8°C. Relative temperature changes were calculated according to $$$\Delta T_{PD}=\frac{(WARM-COLD)}{0.5*(WARM+COLD)}/m_{PD}$$$ (Eq.2) with m_PD≈-0.003 % /°C. T₁ saturation effects were considered for by computing β from Eq.1 based on the ZTE data with varying flip angles and calculating the relative temperature change from β-corrected M₀.

Figure 2 shows the calculated ZTE signals for ideal PD-weighting (β<<1), a PD-weighting achieved with α=0.5° for bone (T₁ = 220ms) and muscle (T₁ = 1.6s) plotted against ΔT assuming a TR=0.708ms. Receiving with the body coil resulted in SNR of 44.95 (brain) and 20.30 (bone); a SNR of 94.22 (brain) and 56.16 (bone) was obtained when receiving with the head Tx/Rx coil (Figure 3). Overall, the head coil provides a 2.1 and 2.7 fold increase of SNR in brain and bone respectively over the body coil. This translates to an accuracy of 3°C (brain) and 3.5 °C (bone) when using the head coil. Figure 4(a) shows a T₁ map at room temperature extracted using ZTE images (α=1°/4°) used for β-correction. Figures 4(b) and 4(c) show temperature maps of an axial slice of the T₁/T₂ phantom calculated from PD-weighted ZTE images without and with T₁-correction, respectively. A RMSE of 24.5% (without T₁-correction) could be reduced to 1.5% with T₁-correction applied (example in probe #2, T1≈220ms) (Figure 5).

ZTE is known to most effectively achieve PD weighting. It has high spatiotemporal encoding efficiency, is robust to B₀ and chemical shift off-resonance effects, and is insensitive to eddy currents. Radial sampling schemes are known to be motion-insensitive while acquisition, but motion in between temperature measurements has a critical impact on accuracy and therefore image registration needs to be applied when using ZTE for MR thermometry. Experiments performed here show that there are two primary contributions to ZTE based thermometry accuracy: (i) SNR, and (ii) correction for variability in ΔT₁/°C. High SNR is crucial for accurate temperature imaging, although the most critical impact of this work shows that ZTE based thermometry measurements must take into account and correct for contamination from T₁ saturation effects and it is expected to further enhance accuracy through B₁-correction.

To conclude, ZTE based MR thermometry with T₁-correction is a promising technique for temperature quantification in muscle tissue and bone.