RF engineers and clinicians interested in safety aspects at UHF, in particular in the application on pediatric populationSubject-specific SAR measurements are not available in current MR systems; simulations must be performed for RF fields and SAR analysis. We combine SAR simulations performed on generic anatomic models with subject-specific measured B₁⁺ maps, predicting local and global head SAR.

Electromagnetic simulations. We used the Finite Integration Technique (FIT) in the CST MW Suite. We simulated a ¹H 298 MHz quadrature birdcage head coil (Nova Medical, Wilmington, MA, USA), loaded by a human head model derived from the 2×2×2 mm³ voxel-size anatomic model Ella (adult, 59 kg), Virtual population, ITIS foundation (Fig. 1). The maximum local (10g) and global SAR [W/kg] for the two following MR sequences were calculated: 1) Axial “Zero” Time-of-Echo (ZTE) sequence¹ (“SILENT”), 384 FA=4° hard pulses of 12 usec length, 5 inversion and saturation pulses per TR (TR=525 msec); 2) Sagittal FLAIR², 240 FA=120° hard pulses of length 336 usec per TR (TR=8 sec). The ratio rSAR between maximum local and global SAR was computed. Next, we simulated various model positions (rotating the head at 15°,30°,45° from the original position on z-axis, moving the head of ±1cm along the 3 axes) recording SAR and rSAR worse case. Finally, we repeated the procedure with a human head derived from the anatomic model Billie (child, 35 kg), Virtual population, ITIS foundation. To allow comparison with phantom experiments, simulations were run with the coil loaded with a cylindrical phantom (Table I).

Measurements. For 19 adults and 27 children: we collected age, weight. We acquired |B_1,map⁺| with a Bloch-Siegert³ centred on slices corresponding to those used in simulations: parameters TR=33 ms, TE=15 ms, RBW=15.6 kHz, thk=3.5mm, matrix-size 64x64, square FOV 22 cm, 2 nex (total acquisition time: 9 s). Measurements were acquired on a GE MR950 7T human system (GE HealthCare, Milwaukee,WI, USA) using the coil equipped with a 32-rx array. For each slice where |B_1,map⁺| was measured, a coefficient C, proportional to avg(|B1,map⁺|) /B_1,nominal⁺ was calculated and used to scale the SAR simulated on the anatomic models; selection between Ella or Billie was performed on subject-weight basis.

Phantom experiment. The cylindrical phantom (Table I) was created as in ⁴. ΔT generated by SILENT (duration: 462 s) was measured with a fiberoptic temperature probe (Neoptix Canada LP, Quebec, Canada, resolution: 0.1 °C) placed at the location of anticipated maximum local SAR. Before the measurement, the phantom was equilibrated to room temperature; thermal insulation was achieved as in⁵. By applying heat equation, temperature-based measured SAR can be determined⁵. The temperature sensor has length of 2 cm; since the mass of 2 cm³ of agar solution is 10 g, we can assume that temperature-based measured SAR is 10g SAR. We acquired |B₁⁺| in the axial slice crossing the center using the Bloch-Siegert sequence; the coefficient C was calculated and used to scale the simulated SAR for SILENT.

The ΔT generated in the phantom by SILENT measured by fiberoptic temperature probe was 0.3 °C; the temperature-based measured SAR was 2.74 W/Kg. The predicted phantom maximum local SAR (10 g) for SILENT sequence was 3.1 W/Kg. Table I summarizes minimum, maximum, average and standard deviation of the age, weight, predicted global SAR. Predicted maximum local SAR (Fig 2) is calculated by multiplying the global SAR by rSAR (3.4 in Ella, 3.2 in Billie).

There is a good agreement between temperature-based measured SAR and the predicted one. SILENT and FLAIR can be safely used in all subjects, both adults and children; limits on maximum local (10 W/kg) and global (3.2 W/kg) SAR are met. FLAIR is more SAR demanding than SILENT. Predicted SAR exposure does not show a significant variation with subject weight. The trend towards higher SAR exposure as subject weight increases was first shown at 1.5T in⁶, where global SAR measurements were reported and a theoretical explanation was provided. However, the theoretical explanation was derived under the assumptions of a homogenous cylindrical/spherical load with an externally homogeneous magnetic field and that the maximum peak power occurs at the surface of the load; these assumptions are not valid when using a 7T birdcage coil loaded with heads. At 7T the fields generated by the coil inside the heads are inhomogeneous, and this leads to inhomogeneous |B₁⁺| and to hot spots which can occur also far from the surface⁷. During a sequence, the system set the power to obtain the desired |B₁⁺|; it follows that the corresponding SAR can violate the linear increase with subject weight⁷.