For MR guided focused ultrasound treatments in or close to bone, such as for transcranial applications focusing through the intact skull bone and treatments of bone metastases, significant heating can occur in the bone. MRI of bone is in general challenging due to the short T2 of bone. For MR temperature imaging the short T2 also severely decreases the accuracy of the standard proton resonance frequency shift method. Instead researchers have investigated the temperature dependence of the MR signal intensity (SI) and T1 relaxation time for temperature monitoring^1–4. Miller¹ and Fielden³ et al showed that the SI from cortical bone decreases with increasing temperature using ultrashort echo time (UTE) pulse sequences, and Ramsay² et al found that, contrary to what Miller and Fielden observed, the SI increases with increasing temperature using a short TE gradient recalled echo (GRE) pulse sequence. Han⁴ et al further showed that T1 increases with temperature, also using UTE.

In this work we investigate the temperature dependence of the SI (dSI/dT) and the T1 and T2* relaxation times (dT1/dT and dT2*/dT, respectively) using 3D UTE and short TE 2D GRE to investigate which parameter has the highest sensitivity to temperature change.

All imaging was performed on a 3T MRI scanner (MAGNETOM PrismaFit, Siemens Healthcare, Erlangen, DE) using a 3D UTE and a 2D GRE pulse sequence. The UTE sequence utilizes radial, ramped sampling of k-space in 3D starting at the k-space center after a 80 μs hard RF pulse, allowing TEs down to 50 μs. T2* was measured by an exponential fit to data acquired at TE=50, 90, 130, 170, and 250 μs (other scan parameters are listed in Table 1). The 50 μs TE was also used for dSI/dT calculations. T1 was measured using the variable flip angle (VFA) method⁵ with FA 8 and 36° (Table 1).

The spoiled GRE sequence utilizes an asymmetric echo to allow TE down to 1.20 ms. For T2* measurements 6 contrasts with TEs between 1.20 and 7.50 ms were acquired, and then repeated 4 more times with TEs starting at 1.30, 1.40, 1.50, and 1.60 ms, Table 1. The 1.20 ms echo was used for dSI/dT calculations. T1 measurements were performed using the VFA method with FA 8 and 36°.

An approximately 4-cm long bovine femur bone (marrow and connective tissue removed) was placed in a phantom holder that allowed heated water to circulate around the bone, Figure 1. 1 fiber optic probe measured the water temperature, and 3 probes were inserted in 1-mm diameter, 2-cm deep, drill holes in the bone to measure the temperature of the bone. The water was warmed to 4 temperatures (~22, 35, 50, and 65 °C) and the data was collected when all 4 probes measured within 1 °C. The whole setup was places in a 20-channel RF head coil.

Figure 2 shows 2D maps of T1 and T2* relaxation times for the 4 different temperatures for the UTE and GRE scans. Mean and standard error values from a 9x9 ROI close to each probe is shown in Figure 3, together with calculated changes in %/°C. From the UTE data a decrease in SI of 0.3-0.5 %/°C, and increases in T1 and T2* of 0.5-0.9 %/°C and 0.6-0.9 %/°C, respectively, was observed. From the GRE data a slight increase in SI of 0.06 %/°C, and decreases in T1 and T2* of 0.4-0.5 %/°C and 0.3-0.4 %/°C, respectively, was observed.

Using the temperature dependent spoiled GRE signal equation⁶ and the observed values for dT1/dT and dT2*/dT, dSI/dT can be closely predicted for the UTE case, whereas the GRE case predicts a small decrease in dSI/dT.

The decrease in SI for UTE, and increase in SI for GRE, is in accordance with previously published results. However, we observed a lower change in the GRE SI (0.06%/°C) than the 0.9 %/°C reported by Ramsay. The measured change in T1 using UTE agrees well with the 0.6 %/°C reported by Han.

The lower dSI/dT compared to Ramsay may be partly explained by our slightly longer TE (1.20 versus 1.05 ms) which will detect more long T2-component protons. For both UTE and GRE the effect of T1 and T2* on SI are counter-acting each other (both increasing for UTE, and both decreasing for GRE), which reduces the sensitivity of dSI/dT. This may suggest that dT1/dT and dT2*/dT are more suitable candidates for bone MR thermometry, and Figure 3 also shows higher sensitivity for relaxation times that for SI.