Evaluation of specific absorption rate and heating in children exposed to a 7T MRI head coil

Abstract

Purpose: To evaluate specific absorption rate (SAR) and temperature distributions resulting from pediatric exposure to a 7T head coil.

Methods: Exposure from a 297-MHz birdcage head transmit coil (CP mode single-channel transmission) was simulated in several child models (ages 3-14, mass 13.9-50.4 kg) and one adult, using time-domain electromagnetic and thermal solvers. Position variability, age-related changes in dielectric properties, and differences in thermoregulation were also considered.

Results: Age-adjusted dielectric properties had little effect in this population. Head average SAR (hdSAR) was the limiting factor for all models centered in the coil. The value of hdSAR (normalized to net power) was found to decrease linearly with increasing mass (R² = 0.86); no equivalent relationship for peak-spatial 10g averaged SAR (psSAR_10g ) was identified. Relatively small (< 10%) variability was observed in hdSAR for position shifts of ±25 mm in each orthogonal direction when normalized to net power; accounting for $B_1^{+}$ efficiency can lead to much larger variability. Position sensitivity of psSAR_10g was greater, but in most cases hdSAR remained the limiting quantity. For thermal simulations, if blood temperature is fixed (i.e., asserting good thermoregulation), maximum temperatures are compliant with International Electrotechnical Commission limits during 60-min exposure at the SAR limit. Introducing variable blood temperature leads to core temperature changes proportional to whole-body averaged SAR, exceeding guideline limits for all child models.

Conclusions: Children experienced higher SAR than adults for the 297-MHz head transmit coil examined in this work. Thermal simulations suggest that core temperature changes could occur in smaller subjects, although experimental data are needed for validation.

Keywords: 7T MRI; RF safety; electromagnetic simulation; pediatric imaging; thermal simulation.

FIGURE 1

(A) Illustrations of Nina and Eartha child models positioned with brain centered within head coil. (B) Scattering parameter S₁₁ for all models. (C) Scattering parameter S₁₂ for all models. Note that S₂₁ is identical to S₁₂, and S₂₂ was very similar to S₁₁ in all cases, so they are not shown here. The coil was tuned and matched at 297 MHz (indicated by black line) using the adult model Duke. The resonance frequency shifts down when loaded with the child models; the light gray box indicates ±1% in frequency. The coil was not retuned for each model; – 297 MHz was used for all data presented

FIGURE 2

(A) Head average specific absorption rate (hdSAR) and peak spatial SAR (psSAR_10g) as a function of model mass. Two normalizations commonly used for SAR estimation on commercial scanners are plotted ‐ to net power (ie, forward‐reflected) and to average ${\mathrm{B}}_1^{+}$. Linear best fits for hdSAR versus mass are also plotted (dashed lines; see Equations [4] and [5]). (B) Ratio of psSAR_10g to hdSAR for each model. A ratio below 3.125 indicates that the limiting value is the hdSAR (International Electrotechnical Commission [IEC] limit 3.2 W kg⁻¹) as opposed to psSAR_10g  (IEC normal mode limit 10 W kg⁻¹); all models are in this regime

FIGURE 3

The ${\mathrm{B}}_1^{+}$ and 10g averaged SAR (SAR_10g) for all models, depicted at the same spatial scale. Top row: ${\mathrm{B}}_1^{+}$ distributions in central axial plane for mean ${\mathrm{B}}_1^{+}$ = 1 μT; this slice is the one used to normalize the SAR distributions also reported in Table 4. Upper row: ${\mathrm{B}}_1^{+}$ in central transverse section; middle/bottom rows: SAR_10g distributions as maximum projections in sagittal/coronal views, respectively

FIGURE 4

Changes in simulated SAR for all models shifted inside the coil. In each case, results are given for both the SAR normalized to net power and to ${\mathrm{B}}_1^{+}$ (the latter are shown with transparent bars). In this case the ${\mathrm{B}}_1^{+}$ was measured in the same anatomical slice (ie, it shifted with the subject). Typically the change in SAR normalized to ${\mathrm{B}}_1^{+}$ was greater than that normalized to net power, if no transparent bar is visible that indicates the opposite was true. Note that the y‐axis scales for the bottom row are different than the other plots, as superior–inferior (SI) shifts lead to much larger changes. Abbreviations: AP, anterior–posterior; LR, left–right

FIGURE 5

Summary of results from thermal simulations, run at power level and resulting in hdSAR = 3.2 W kg⁻¹ for each model. Top row: Results for fixed blood temperature; bottom row: results for variable blood temperature. Solid lines in (A) and (B) indicate overall maximum temperature, and dashed lines indicate core temperature; the latter is defined as the average over the heart and brain. (A) With fixed blood temperature, after 1 hour of exposure, the maximum temperature does not exceed 39°C and core temperature increases are about 0.1°C, for all models. (B) When blood temperature is variable, IEC guidelines for core (change of 0.5°C) and/or maximum local temperature (39°C) are exceeded for most models. (C) Times to exceed IEC guidelines for the data presented in (B). Asterisks indicate that the limit was not exceeded. In general, for the smaller models, core temperature limit is exceeded before maximum temperature limit

FIGURE 6

Maximum projections of temperature after 60 min exposure at hdSAR = 3.2 W kg⁻¹ for thermal calculations with fixed blood temperature (top two rows) and variable blood temperature (lower two rows). Within each block, the upper row is a sagittal projection, and the lower row is coronal. Fixed blood temperature leads to generally lower temperatures than when variable blood temperature is modeled, but the spatial distributions are similar