Introduction

MR thermometry provides noninvasive temperature measurements in near real-time ¹. After implantation of DBS leads, magnetic resonance imaging (MRI) may be clinically necessary. However, heating induced near lead electrodes during MRI remains a safety concern ². An MR thermometry scout may enable heat monitoring and introduce cooling periods to keep heating below a safe threshold and ensure safety.
In this work, we demonstrate the feasibility of using MRF for temperature measurements near DBS lead by exploiting T₁ relaxation time’s dependence on temperature while simultaneously estimating B₀ map.

Method

MRF implementation
The MRF sequence uses parameters reported in previous literature ^{4, 5}. The acquisition time was 18 seconds/slice. Data were reconstructed using singular value decomposition (SVD) dictionary matching ⁶. The reconstruction time was ~30 seconds/slice.

Calibration experiment
Figure 1(a) shows the setup for calibration experiment. We ran MRF sequence every 5 minutes after experiment started. Fluoroptic probes measured temperatures every second. The experiment lasted 125 minutes.
Four ROIs with 60 voxels were selected based on the position of probes (Figure 2(a)). The mean T₁ of each ROI was calculated. The corresponding probes’ temperatures were averaged over MRF acquisition (18 seconds). Simple linear regression and R² statistics were calculated using least-squares fitting.

Lead experiment
Figure 1(b) shows the setup for lead experiment. We ran MRF sequence first to measure baseline temperature, followed by a 16:26 (min:s) turbo spin-echo (TSE) sequence to induce heating. After the heating, we ran MRF protocol every minute for 10 minutes to measure temperature as it returned to baseline. Fluoroptic probes measured temperatures every second. We repeated the cycle three times and acquired a TSE structure image for ROI selection. The experiment lasted 75 minutes.
One ROI with 60 voxels was selected 1 mm away from the tip of the lead. The maximum value of ROI was calculated. The corresponding probes’ temperatures were averaged over MRF acquisition.

B₀ experiment
Due to orientation limitations for gold standard B₀ maps acquisition, we conducted a separate experiment to investigate the effects of B₀ inhomogeneity around lead. The lead was inserted into the cold bovine muscle, same as lead experiment. A vendor sequence (fast GRE-based B₀ mapping) with TR=250 ms and FA=30^o was run to acquire gold standard B₀ maps. The acquisition time was 2:12 (min:s). We then ran MRF sequence to compare B₀ maps.

Spatial temperature gradient analysis
We selected ROIs 1 mm, 3 mm, and 6 mm away from the tip of lead electrode to investigate the spatial temperature gradients near the lead.

Results/Discussion

Figure 2(a) shows that T₁ decreases as the heated bovine muscle cooled down for the first 30 minutes of the experiment. The T₁ of the room temperature water bottle serves as a reference and stays ~2 seconds for entire experiment. The cold muscle has longer T₁ than heated muscle because the heated muscle loses water during heating process which we drained. Figure 2(b) presents the best fit line for the center of cold muscle and center and periphery of heated muscle and their respective R² statistics. All R² values are larger than 0.988, indicating a strong linear relationship between temperature and T₁. This linear relationship agrees with previous literature ^{1, 7}.
Figure 3(a) and (b) show the T₁ and temperature map of the first cycle in the lead experiment. T₁ increases as temperature increases due to TSE-induced heating. The heating was more evident around the lead. An ROI was selected in (c) to investigate the temperature estimated by MRF and temperature measured by fluoroptic probes. (d) shows that MRF estimated temperature has a similar trend as probe temperature, but the range is smaller compared to the probe temperatures. This may be because the ROI was 1 mm away from the tip of the lead and the spatio-temporal averaging of the T₁ caused the ROI to experience less amount heating as the probe.
Figure 4 shows that the B₀ map estimated using MRF had similar regions of inhomogeneity but underestimated gold standard B₀ map. The gold standard B₀ map showed ~200 Hz inhomogeneity around lead area, while the MRF estimated ~150 Hz. Figure 4(b) presents joint T₁ and B₀ mapping. The figures show that more signal was recovered resulting in a longer T₁. This may have a significant impact on accurate estimation of temperature near lead because of the inhomogeneity generated by lead electrodes.
Figure 5 shows the positions of three different ROIs that are 1 mm, 3 mm, and 6 mm from the tip of the lead and their maximum values for each time point respectively. The 1 mm ROI has the largest temperature change (~ 10 degrees) and the 6 mm ROI has the smallest temperature change (~ 4 degrees), similar to findings in ². The starting temperatures for ROIs were different because the temperature of the muscle was not homogeneous when the experiment started.

Conclusion

We investigated the feasibility of measuring temperature near DBS lead using MRF based T₁ and B₀ mapping. T₁ and temperature show strong correlation (R² > 0.988). We observed a maximum change of 10.2 ^oC near leads during calorimetry experiments, validated by fluoroptic probes. The effects of B₀ and distance from the lead were investigated.

Acknowledgements

This work was supported, in part, by GE-Columbia research partnership grant and also performed at Zuckerman Mind Brain Behavior Institute MRI Platform, a shared resource, and Columbia MR Research Center site. The subject matter, content, and views presented do not represent the views of the Department of Health and Human Services (HHS), US Food and Drug Administration (FDA), and/or the United States.