Synopsis

Keywords: Thermometry, Phantoms

Gradient Echo Proton Resonant Frequency (GRE-PRF) based thermometry, a standard approach to temperature mapping, was assessed at 0.5T. Experiments were performed using a 3 slice GRE acquisition at a resolution of 2x2x5mm and an update rate of 7.7s. Phantom and in-vivo measurements of the temperature stability yielded uncertainties of 0.78°C and 1.48°C respectively. Furthermore, a cooling experiment with the phantom showed excellent agreement to temperature changes simultaneously measured with a temperature probe. All results suggest that GRE-PRF thermometry is feasible at 0.5T.

Introduction

Mid-field MR imaging is an active area of study [1-2] with renewed interest in implementing on these systems a variety of imaging methods usually reserved for higher fields. One such implementation is temperature mapping, which allows for non-invasive tracking of thermal therapies [3]. The standard approach to MR thermometry is the gradient echo proton resonance frequency (GRE-PRF) method. GRE-PRF thermometry in the mid-field is difficult as the longer T2* values cannot be optimally leveraged without comprising update rates or spatial resolution. In this work, we characterized GRE-PRF thermometry through a series of phantom and in-vivo stability and cooling experiments.

Methods

Phantom Production
A custom double nested cylinder was produced, with the outer layer acting as a reference region, and the inner cylinder resting in polyurethane foam susceptibility matched using Pyrolytic Graphite in accordance with [4]. Schematic of the phantom is shown in Figure 1A. Hydroxyethylcellulose (HEC) was used as a tissue mimicking gel inside of the phantom, produced using the method described in Annex L of [5], for the solution with 96.85% water, 3% HEC [Sigma-Aldrich; St. Louis, Mo], and 0.15% NaCl, giving a relative dielectric permittivity of 78, conductivity of 0.47 S/m, density of 1001 kg/m. The HEC gel was then doped using Copper (II) Sulfate [Sigma-Aldrich; St. Louis, Mo] to a 2mM concentration to produce a T1 relaxation time similar to brain tissue (635 ms). The alpha parameter required for PRF thermometry was calculated by simultaneously measuring the center frequency and the temperature as a sample of heated HEC gel cooled.
Temperature Stability Measurement
Temperature stability imaging was performed on the phantom described above and on a healthy volunteer with informed consent in compliance with health and safety protocols. All imaging was performed using a 0.5T head-only MR scanner [Synaptive Medical, Toronto, Ontario] equipped with an 8 channel head-coil. In both cases, 64 repetitions of a 7.7s, 3-slice, GRE-PRF acquisition was run (TE = 23.75ms, TR = 86ms, BW= 20kHz, matrix size = 110x90, slice thickness = 5mm, slice gap = 5.5mm, FOV = 22x18cm, flip angle = 28°).
Temperature Tracking Experiment
To validate the ability of tracking temperature change using the PRF method, the inner container in the phantom was heated in a hot water bath to ~60°C, then a fibre optic probe was inserted into the HEC gel acting as a thermometer, with tracking done on a NOMAD-Touch [Qualitrol; Fairport, NY] system (±1 °C; ±0.1 °C Resolution, -20 to 80 °C range, calibrated using a monitored hot water bath). The heated sample was placed in the phantom and a series of 702 repetitions of the GRE-PRF acquisitions described above were collected.
Temperature Map Reconstruction and Data Analysis
Phase differences were calculated relative to the first phase image acquired. A first order background phase correction was applied to the data based on the reference region in the phantom, and the outer rim of brain in the in-vivo case. This phase difference was converted to temperature, with the calibrated alpha parameter for the HEC gel and –0.01 ppm/°C for in-vivo measurements [3]. For the stability measurements, voxel-wise temperature uncertainty was calculated as the standard deviation of the temperature measurements over the time-course.

Results

A schematic of the experimental set up, along with determination of the alpha parameter for HEC gel is shown in  Figure 1, with α = (-8.74 ± 0.12) x 10^-3 ppm/°C.
The temperature uncertainty over 64 GRE-PRF acquisitions is shown in Figure 2, with average value of 0.78°C. In Figure 3, cooling of heated HEC gel is shown using GRE-PRF and the temperature probe. The PRF temperature shown is an average of a 4mmx4mm region near the probe. In-vivo temperature uncertainty for the GRE-PRF sequence is displayed in Figure 4. The average uncertainty over the slice shown is 1.48°C.

Discussion

Sub 1°C phantom temperature uncertainty over ~8 minutes of scan time shows the potential of temperature monitoring using this method, and as seen in Figure 3, the GRE-PRF method was indeed effective at tracking temperature. The TE used by this sequence is well below the optimal value of T2* at 0.5 T; however, average in-vivo temperature uncertainty was still below 2°C, providing sufficient precision. Future work will be done to fully characterize the change in temperature precision with echo time both for the phantom used in this study and in-vivo.
A benefit to approaching temperature mapping using a relatively simple sequence such as GRE is the ease of implementation in a variety of scanners. With good matching to tracked temperature changes, and low uncertainty when imaging parameters are optimized, there is a strong case for its use as a standard method for PRF thermometry. A temporal resolution of 7.71s is adequate; however, further reduction using GRE with relatively long TE is difficult. For increased update rate, an EPI based PRF approach is likely a promising alternative.

Conclusion

This study has encompassed the development and validation of GRE-PRF temperature mapping on a 0.5T MR system. The method was able to accurately measure temperature in a cooling experiment, providing adequate temperature uncertainty in vivo with clinically acceptable update rates, spatial coverage, and image resolution.

Acknowledgements

No acknowledgement found.