MRI scanner‐independent specific absorption rate measurements using diffusion coefficients

Abstract Objective The purpose of this study was to measure specific absorption rate (SAR) during MRI scanning using a human torso phantom through quantification of diffusion coefficients independently of those reported by the scanner software for five 1.5 and 3 T clinical MRI systems from different vendors. Methods A quadrature body coil transmitted the RF power and a body array coil received the signals. With diffusion tensor imaging, SAR values for three MRI sequences were measured on the five scanners and compared to the nominal values calculated by the scanners. Results For the GE 1.5 T MRI system, the MRI scanner‐reported SAR value was 1.58 W kg‐1 and the measured SAR value was 1.38 W kg‐1. For the Philips 1.5 T MRI scanner, the MRI system‐reported SAR value was 1.48 W kg‐1 and the measured value was 1.39 W kg‐1. For the Siemens 3 T MRI system, the reported SAR value was 2.5 W kg‐1 and the measured SAR value was 1.96 W kg‐1. For two Philips 3 T MRI scanners, the reported SAR values were 1.5 W kg‐1 and the measured values were 1.94 and 1.96 W kg‐1. The percentage differences between the measured and reported SAR values on the GE 1.5 T, Philips 1.5 T, Siemens 3 T, and Philips 3 T were 13.5, 6.3, 24.2, 25.6, and 26.6% respectively. Conclusion The scanner‐independent SAR measurements using diffusion coefficients described in this study can play a significant role in estimating accurate SAR values as a standardized method.

due to the deposition of RF power into the body, and this is a significant safety concern. 1-11 It is thus necessary to determine the RF energy absorbed by the body in terms of the specific absorption rate (SAR). According to International Electrotechical Commission (IEC), the SAR value should be limited to 3.2 W kg -1 for the head and 4.0 W kg -1 for body applications for durations of 6 min. 12 Similarly, the Food and Drug Administration (FDA) of the United States requires that the SAR should be less than 4 W kg -1 when averaged over the entire body for 15 min and 3 W kg -1 for the head for 10 min. 13 The risk of hyperthermic tissue damage is relatively serious for neonates and for children who cannot communicate verbally, as well as for patients who have insensate limbs and those who are under anesthesia during the MRI scan.
Commercial MRI scanners provide an estimated SAR level for each scan; this level is calculated from the RF waveforms and sequence parameters, system calibration, Q factors and loading of the RF transmit coil, etc. The SAR calculation assumes certain average parameters, which in reality can vary from scanner to scanner and may change over time. 14 Incorrect manufacturer-reported SAR values have been acknowledged for clinical MR imaging systems. [14][15][16] For example, one study found a scanner overestimated the SAR by up to 2.2 folds. 15 Even before the highest allowed SAR level has been reached, a patient's sweating during an MRI can raise concerns of possible overheating. On the other hand, overestimating the SAR can prevent certain important scans to be run on a patient. It is also conceivable that a malfunction in the quadrature RF transmit coil can generate RF with higher levels in the counter rotating component, resulting in higher than expected power deposition levels.
Direct estimation of SAR values independent of the level calculated by MRI scanners is therefore desirable.

2.B | Human torso phantom morphology
A cylinder-shaped human torso phantom (50 cm (L) 9 43 cm (W) 9 28 cm (H)) was constructed on the basis of U.S. anthropometric reference data 25 (Fig. 1). The airtight plastic phantom container T A B L E 1 Image acquisition parameters at 1.5 and 3.0 T.  2.C | Independent SAR assessment using diffusion measurement Four optic fiber temperature sensors (OFS, Neoptix Inc., Quebec, Canada) were placed at the periphery of the gel phantom at 28°C to certify that there was minimal heat loss to the environment during the measurements (Fig. 2). We measured the initial temperatures of the sensors positioned in the phantom and the time it took to reach equilibrium with the environment. We considered that thermal equilibrium has been reached when the difference between temperatures measured by the sensors in the phantom and temperature inside the magnet bore was less than 0.1°C.
On each MRI scanner, the heating of the gel phantom caused by a high SAR sequence was assessed by the changes in the mean diffusivity (MD) value, before and after running the high SAR image sequence. 26 A region-of-interest (ROI)-based MD calculation was performed for the SAR measurements. In the ROI-based quantification, the average signal intensity within the ROI as shown in Fig. 3 for the b = 0 image and each high b diffusion-weighted image was measured first. Using these values, a diffusion tensor was calculated, and MD value was obtained. The water diffusion coefficient in the gel was practically identical to that of free water, 27 and the diffusion coefficient (D) was very sensitive to the temperature (T). 28,29 The temperature was calculated using the following equation: 28 where D 0 = 1. Verification of temperature changes obtained by the diffusion coefficients using Eq. 1 was performed by comparison to those measured by four optic fiber temperature sensors positioned as in Fig. 2.
The mean diffusion coefficients within each ROI which were F I G . 1. Phantom morphology mimics the shape of the human torso.
F I G . 2. Location of four optic fiber temperature sensors in the phantom periphery, used to measure the initial temperatures of the phantom and the time taken to reach equilibrium with the environment.
F I G . 3. The apparent diffusion coefficient within an ROI (red) which was manually placed on the periphery of the phantom was quantified using a set of diffusion-weighted images.
manually drawn around the temperature sensors were calculated using software written in IDL 8.4 (IDL Research Systems Inc., Boulder, CO, USA) before and after the high SAR image sequence at 3 T.
The MD value was the average over 14 pixels within the ROI in one slice showing the tip of the sensors.
For each study, the phantom was placed in the scanner room for at least 24 hr to establish thermal equilibrium with the environment.
The same phantom weight (18 kg) was entered into the MRI system at registration. First, the SAR value induced by the diffusion tensor imaging (DTI) scan was measured for the torso phantom on each MRI scanner using repeated DTI scans. Second, in order to measure the SAR value caused by the high SAR sequences, an axial DTI scan was initially conducted, followed by several minutes of high SAR scan. The DTI scan was then repeated. The scanner-specific DTI acquisition parameters are listed in Table 2. Auto-shim was utilized in all studies. The mean diffusivity within a ROI which was manually placed at the periphery of the phantom was evaluated for each DTI scan using software written in IDL 8.4 (Fig. 3), and the temperature change derived from the difference between the MD maps was estimated (Fig. 4). The standard error of the mean diffusivity for the ROI ranges from 0.0005 9 10 -3 to 0.0006 9 10 -3 mm 2 s -1 for the 5 scanners based on repeated measurements.
The SAR values and the temperature change are related according to the equation below 26 : Here, C p (= 4.18 kJ/(kg°C)) is the specific heat of the phantom, DT is the temperature change in°C, and TA is the acquisition time of the pulse sequence. The first term on the right-hand side is the heating from the DTI sequence, and the second term on the same side represents the heating from the high SAR sequence under investigation.
One SAR measurement was performed in one MRI session, and then the phantom was placed in the scanner room for one day to reach thermal equilibrium with the environment. SAR measurements were repeated at least 10 times for each MRI scanner and the mean and standard deviation (SD) were calculated. Percentage differences between the measured mean and reported SAR values on the MRI scanners were calculated.

| RESULTS
The results of the temperatures measured by the optic fiber sensors positioned at the phantom periphery and the times taken to reach equilibrium with the environment are summarized in Table 3. At initial phantom temperatures of 28°C, it took more than 4 hr to reach equilibrium with the environment. The sign '-' means that measured SAR value was lower than the reported one.
T A B L E 3 Phantom temperatures (mean AE SD) measured via optic fiber sensors and times to equilibrate with temperature inside a magnet bore at 3 T. tion of this work is that a SAR measurement based on diffusion in a "torso phantom" is not the same as a human measurement, which will also include the effects of perfusion, and sample heterogeneity.
Our results confirm that some vendors' SAR values may not be reliable, and can only be used as an approximate guide. Although our study was done using 1.5 and 3 T scanners, the same approach can be applied to higher B 0 fields. The scanner-independent SAR measurements described in this study using diffusion coefficients thus can play a significant role in estimating accurate SAR values as a standardized method. This study can give radiologists greater confidence when they scan patients clinically. In addition, this test can be used as a tool for quality assurance and the calibration of MRI systems.

| CONCLUSIONS
We demonstrated that SAR values measured by quantification of the water diffusion coefficient can be used as a feasible alternative to that calculated by clinical MRI systems.

This work was funded by the National Research Foundation of
Korea (NRF-2014R1A1A2054037).