Figure 3: maximum scattered E-field at the tip of a metallic wire plotted against wire length. The incident field was a uniform (phase/amplitude) plane wave of 1 V/m. Different transmit frequencies were considered (21-298 MHz) as well as insulated (dashed) and bare wires (solid).

Figure 4: violin plots showing, for a wide range of investigated positions and trajectories, the distributions of induced tip current for straight wires and realistic DBS implants (a and c normalized to head SAR, b and d normalized to B1). Different implants lengths, insulation types and transmit frequencies were considered.

Figure 1 – Numerical models of a 1.5 T 16-rung low-pass birdcage coil tuned at 64 MHz, a 3 T 16-rung low-pass birdcage coil tuned at 127 MHz, a 7 T 16-rung hybrid birdcage coil tuned at 297 MHz, and a 10.5 T 8-channel bumped dipole coil tuned at 447 MHz all loaded with a uniform human head model with no implant. The geometrical dimensions of the coils were the same as the coils reported in the literature^2-5.

Figure 3 - 1g-SAR distributions in patient 10 (ID10) for the 1.5 T, 3 T, 7 T, and 10.5 T coils on an axial plane passing through the electrode contacts. The coils' input powers were adjusted (A) to generate an average of B₁⁺ = 2 μT and (B) to produce a GHSAR = 3 W/kg over an axial plane passing through the patient's eyes.

Figure 2. (a) A schematic from the simulation environment showing the overall simulation volume (white box), the tissue volume (purple), the dipole antenna (yellow), and the high mesh density volume around the screw (dark purple). (b) Representative example of a SAR map, showing the location where the SAR_max value was typically located at the screw tip (arrow).

Figure 3. Graph of the SAR_max for each field strength configuration as a function of implant length embedded in bone tissue. The simulated input RF power levels were normalized to ensure the same transmitted power of 1W by the dipole antennas across all field strengths.

Figure 1; A-C) computer-aided designs (CAD) of the RF transceiver representing sections of the RF array with the position of the phantom and implant inside the phantom. D) manufactured RF transceiver, phantom and power dividers.

Figure 4; Simulation results of coronal unaveraged specific absorption rate (SAR) maps of a plane through the center of the implant using degenerate birdcage mode (CP), orthogonal projection (OP) and genetic algorithm-based (GA) shimming.

Figure 3 – Maximum temperature increase in the whole body volume after 360 s exposure versus the axial position of the hip implant in the coils, corresponding to the twelve body positions from 1 to 12. Results for EPI, True-FISP, GRE and TSE (with constant dead time) sequences. Dimensions of the prosthesis: stem length 142 mm; hemispherical head diameter 30 mm; acetabular shell diameter and thickness 66 mm and 8 mm; screw length 34 mm; polyethylene liner thickness 10 mm.

Figure 4 – Time-spatial evolution of the temperature increase around the prosthesis for three analysed sequences (screenshots after 50 s, 200 s and 360 s). For the 3 T group, the results for the EPI sequence are not reported, being almost identical to those at 1.5 T. The results for the TSE sequence refer to a constant dead time of 5.67 s. A region of influence around the implant was defined as a parallelepiped box of size 21.7 cm × 18.8 cm × 28.2 cm.

Figure-5:MRI experiments at 3T using the diode and $$$Q_S^E$$$ for pTx mitigation and imaging. The target location for the B₁-shim is marked with a red cross. A-D) Coronal and E-H) axial GRE images for the four different pTx modes WC, CP, OP, and NM. Signal intensity profiles along I) x-axis, J) y-axis and K) z-axis for each excitation mode. L) shows induced diode signals on the implant for these modes. OP and NM show substantially reduced tip signal compared to the B₁-shim (and dramatic reduction compared to WC).

Figure-3: Results for the diode sensor in the mapped area in the testbed. The implant is immersed 120 mm into the phantom. (A) The diode readings for CP, OP and NM. OP and NM are derived from eigenvalues of the acquired $$$Q_S^E$$$. In (B), the heating rates measured by the thermistor are mapped, when the same CP, OP and NM are applied for 2 s per implant position. (C) Diode-signal (i.e. E-field) ratios (logarithmic scale) for each mode, indicating improved or reduced patient hazard. Compared to CP, both OP and NM reduce the E-field nearly everywhere (99%) and by a factor ≥2 in 93% of the positions.

Electrode configurations of ROCA-ROLI(2),ROLI(1)-ROLI(2) and LCIP-LTFU with a table summarised their targeted tissue, brain hemisphere and number of electrodes.

RF shimming for the 16ch loop coil on the whole brain shown in the transverse plane, the 1^st column illustrates the mean B₁⁺ maps and the 2^nd column the MIP of the SAR_0.1g maps a) without electrodes b) with ROLI(2)-ROCA, normalised to 1W total input power.

Figure 1: Schematic of the TF evaluation approach for AICs and AIGWs. A) Hot spot detection with high resolution E field mapping during continuous RF excitation with a dipole antenna. B) TF measurement. C) TF validation by E field mapping with and without the AIGW. D) A detailed schematic of the electrooptical sensor. E) Dipole antennae used for generating uncorrelated incident fields.

Figure 2: Results of the hotspot detection measurements for 100cm, 80cm and 60cm insertion lengths from top to bottom. The location of the hotspot remains constant, and it is 4mm before the tip point.

Figure 3: Comparison of the ViP model FATS with the ASTM test-field. E10g maximum intensity projection (MIP) for all imaging positions at 1.5T, normalized to normal operating mode. Maximum E10g in FATS is up to >4 times higher than the ASTM test field of 120 V/m, resulting in a factor >18 in power (red box). For the complete assessment, other models, MR coils, and field strengths have to be considered.

Figure 2: Comparison of the different Tiers: Tier 2, where the total E-field of the passive implant is averaged on 10g cubes. Proposed Tier 2.5, where the tangential E-field along representative trajectories over the 3D volume of the passive implant are averaged. Tier 3, where the 1D tangential E-field is extracted and used as input for the transfer function.

Deposited lead-tip power for the 100 cm long single-wire and double-wire lead (wire distance d =1.3 mm) with main-wire break under isoelectric conditions. Also, (i) 100 cm long single-wire lead with wire break at b = 50 cm and a gap size g = 1, 2, 5 and 10 mm, (ii) 100 cm long double-wire lead with wire distances d = 1.8 and 0.55 mm and the main-wire broken at b = 50 cm, and (iii) 100 cm long double-wire lead with wire distance d = 1.3 mm and both wires broken at b = 50 cm. Experimental validation measurements (m₁ – m₄) of the four 50 cm break cases are shown with slight offset for clarity.

(a) Generic implant with a 100 cm long single-wire lead, with a break of dimension g at distance b from the lead-tip. (b) Generic implant with a double-wire lead (wire separation d) and two electrodes (lead-tips).

Figure 2: (A) Superposition of E_tan distributions along different lead trajectories. Similar E_tan values were observed along the intracranial regions across lead trajectories. (B) Distribution of E_tan (color field) and the incident electric field (arrows) for lead trajectories that demonstrate low and high 1gSAR_max values. (C) A DNN was created to predict 1gSAR_max using the extracranial E_tan values as inputs into the algorithm due to similarities in E_tan values in the intracranial regions.

Figure 4: (A) Performance of network training, as indicated by the mean squared error, for all folds of five-fold cross-validation. (B) Predicted 1gSAR_max values from testing of the DNN plotted against the ground-truth 1gSAR_max values from EM simulations.

Figure 1: A representative image of (a) 1.2T Hitachi Oasis open bore MRI system, (b) RF body coil of a 1.2T Hitachi Oasis open bore MRI system with the Duke human model, (c) RF body coil of 1.2T Hitachi Oasis open bore MRI system with a knee implant in an ASTM gel phantom (65cm x 42cm x 9cm), and (d) detailed CAD model of a knee implant.

Figure 4: Temperature rise contour of a nitinol SFA stent and a knee implant within Duke in 1.5T Siemens Avanto closed bore and 1.2T Hitachi Oasis open bore MRI systems at a scan time of 900 sec.

Figure 1: Contribution of IPG input impedance against the total lead/IPG impedance to 63.87 MHz RF currents at the lead’s Tip electrode. Lead/IPG pairs are MR-conditional labeled for whole-body MRI at 1.5T.

Table 1: Contribution of IPG input impedance against the total lead/IPG impedance to 63.87 MHz RF currents at the lead’s Tip electrode. Lead impedance incorporates the effect of the implantable device’s enclosure in gel slurry. Lead/IPG pairs are MR-conditional labeled for whole-body MRI at 1.5T.

Figure 1: Magnitude and phase of the Biotronik and Abbott leads before and after soaking for over 10 days in isotonic saline solution

Table 1: Characteristic Impedance at 63.87 MHz of pacemaker leads before and after being soaked for over 10 days in saline solution

Figure 5: Maximum intensity projection SAR_1g results are shown near the implantable devices. All the SAR maps were normalized with the B₁⁺ method.

Figure 4: The slice SAR maps with implants are shown in four different imaging landmarks for SAR_1g and SAR_10g. All the SAR maps were normalized with the B₁⁺ method.

Fig. 5: This figure shows the measured temperature rise values after six minutes divided by the square of the incident E-field at each position and the target values from Annex I¹.

Fig. 4: This figure shows the measured E_rms-field distributions, a) E_rms-field at 0° phase, c) E_rms-field at 180° phase and the deviation between measurement and simulation in b) 0°phase and d) 180° phase.

Figure 1: Flowchart with decision tree for MRI examination in patient with a cerebral aneurysm clip, with references to Tables numbers in this ISMRM abstract (which differ in numbering from the full guideline). Zoom in your browser to read. While considering all evidence and considerations it was recognized that withholding an MR exam for large patients groups in which it is likely but not entirely certain that the aneurysm clip is non-ferromagnetic, will impact patient’s diagnostics.

Table 4: Probability that an unknown type of cerebral aneurysm clip is not safe for MRI; as estimated by the working group.

Figure 2. (A) SAR maps and 1-gram local-SAR values for RA and RV leads for the reference position (supine and head-first). (B) SAR maps and 1-gram local-SAR values for a feet-first patient orientation is substantially reduced. (C) Right-left lateral rotations also substantially reduce the 1-gram local-SAR.

Figure 1. Patient position and orientation variations: (A) head-first vs. feet-first; (B) lateral rotation; (C) transverse left-right displacement; and (D) Fowler position. Each was compared to the local-SAR value with respect to a reference position (shaded, supine and head-first) for a thoracic exam.

Figure 2.:The simulation set-up for the EEG traces and sponges. (a) shows the 3D scanned EEG net put on the 29-month-old head mock-up, (b) shows the position of the sponges which is chosen from (a) and used as the reference points for the EEG trace allocation, (c) shows the final drawing of 128 channel EEG traces connected on sponges on the head of the 29-month-old boy voxel model.

Figure 5: EM simulation results. (a) pSAR results displayed in the surface of the 29-month-old voxel model with a 128-channel EEG net, (b) pSAR results in the surface of the 29-month-old without an EEG net (In both cases, the simulation results are normalized to 2µT at the coil center)

Fig 2a: Experimental heating rate of a knee implant #3 at $$$t=0, x=0, y=0$$$ and various $$$z$$$ positions in a clinically relevant orientation. The signal of the central sensor 1 is shown in blue, the lateral in red.

Fig 2b: An axial view of the knee implant with two NTC sensors mounted in sockets that are glued on the implant.

Fig 1a: Comparison of experimental (solid) and simulated (dashed) eddy-current heating of hip implant #1 at scanner coordinates $$$z=-300 \text{ mm}, y=-50 \text{ mm}, x=-145 \text{ mm}$$$. Central sensor in blue, sensors located on the rim in red and green. The EPI sequence was applied from $$$t=0 \text{ to } 20 \text{ s}$$$. The main eddy current circulates along the rim with its large cross section to the vector field $$$dB/dt$$$ of the read gradient z.

Fig 1b: Simulated temperature rise distribution at $$$t=20 \text{ s}$$$. The arrows indicate the sensor positions.

A) HFSS setup with a central test object and in-phase Plane Wave excitation (red dotted), separated by radiation boundaries (blue dotted). Remaining surfaces are terminated by PML boundaries. Outer Box is the overall Simulation Box; Inner Box is the VLD export box. B) close up of the titanium rod tip C) single PW source, D) dual outer PW sources (also used for phase variations), E) three PW sources, F) two inner PW sources and G) four inner PW sources. The 300mm central line is identical with the z-axis.

Root Mean Square of the tuned (150V/m) E-Field Magnitude for the five different excitation setups along a 300mm central line in the centre of the simulation volume. The 100mm tuning line, resp. the test object, is situated between 100mm and 200mm. The dotted red lines indicate the ±1dB (12%) limitation.

Figure 2. Setup for generating artificial tangential electric fields along the lead.

Figure 3. Transfer function of wires with electrode 1 and electrode 4.

SAR maps of the maximum axial, coronal and sagittal slices are shown for Duke and Ella at 0mm, with and without the presence of the EEG cap.

EEG cap and electrodes, shown on the sucrose/saline phantom (Fig.2a). View of the electrodes and the connecting wires on the sucrose phantom in CST. The lumped elements representing 5kΩ resistors are shown. Care was taken to distribute the wires in the same positions as the physical cap (Fig.2b). A similar model from CST showing the EEG electrodes in place on the Ella model. The lumped elements are hidden in this view (Fig.2c).

Figure 3: Measured temperature increase compared to reference of an implant in a phantom of gel (Δ), water (o), or air (X) in the literature (unknown SAR value on the right). The symbol size is proportional to the number of values reported.

Figure 1: Number of implant types for which a certain gradient limit is set by the manufacturer (A) and number a certain SAR limit is set by the manufacturer (B), from MagResoure.

Figure 1: Normalized 3T MRI RF-Induced In-Vitro Temperature Rise at Different Fixed wbSAR and B_1+RMS Levels under Normal Operating Mode for Two Different Cardiac Pacing Leads

Figure 3: Points of Equivalent 3T MRI RF-induced Heating for B_1+RMS vs. wbSAR Scaling under Normal Operating Mode for Two Different Cardiac Pacing Leads

Figure 2 Susceptibility artifacts from the septal occluder, the stent and the endoscopic clip. A bandwidth of 156Hz/pixel is used on the 50 mT system (top row), which is matched by the first 3 T scan (middle row), whilst the bottom row shows the 3 T images acquired with 880Hz/pixel. In the images the direction of the main magnetic field (B₀) and the frequency- and the two phase-encoding directions are indicated.

Figure 1 (a) The setup for the magnetically induced displacement force measurement inside the 50 mT system with the angle of deflection α’, angle of rotation β and absolute angle of deflection α indicated. It is based on the F2052-15 ASTM protocol⁴ with the additional capability of rotating the device to enable measurement of the variation along the z-axis. (b) The central xz-plane of the simulation data for the scalar gradient of the magnetic field, with the measurable area (red) indicated.

A.) shows the measured temperature at the tip of the lead. Only the measurements done with the smaller loop coil are shown, a similar set of measurements were done with a larger loop coil. The slope of the temperature increase at the start of each separate heating test is used to obtain the (measured) SAR at the temperature probe for that specific heating test. The (predicted) SAR that is calculated using the transfer function and the known exposure condition is correlated with the measured SAR showing good agreement for all three leads.

A comparison between the transfer functions obtained with FDTD simulations and the transfer function obtained with MRI measurements are shown. The left plot shows the normalized magnitude of the TF and the right plot shows the phase of the TF.

Figure 1 Distributions of the examination duration for scan regions for Site A, Site B and Site A and B combined.

Table 2 Mean and standard deviation of the examination duration for each scan region across the imaging sites

Figure 2: Gelatin-filled human skull with implanted deep brain stimulation (DBS) leads and fiber-optic temperature sensors, and the assembled phantom: (a) location of the temperature sensors: sensors 1* and 2* implanted inside the skull and sensor 3 at the spiral trajectory; (b) right DBS lead trajectory and temperature sensor 4 and (c) assembled phantom prior to the final fill and temperature sensor 5 [7].

Figure 4: Turbo spin-echo images of the phantom: (a) reconstructed transverse view for implantable pulse generator (IPG) off using transmit/receive birdcage coil; (b) reconstructed transverse view for IPG on using transmit/receive birdcage coil; and (c) a reconstructed sagittal view for phantom illustration purposes using 20 channel receive-only head and neck array coil and body coil for transmission [7].

Figure 1: Senza IPG location based on three different heights indicated as a red circle.

Figure 2: For the specified location of the device (red circle), shown are a) the main magnetic field strength and b) normalized RF voltage squared along the z-axis, which is a measure of (B1+)². Because the RF is negligible at the location of the device, even if the device labeling calls for a T/R head coil, the brain could be scanned with a multi-channel receive head coil, which offers a substantial improvement in image quality.

Figure 2: psSAR-actual is compared with (a-b) psSAR-centre and (c-d) psSAR-9positions. (a-b) The actual peak local SAR was up to 4.6-fold higher than the peak estimated using the centred safety model. The sensitivity of psSAR to positional mismatch increased with the number of spokes but remained fairly consistent across slices (15/21 comparisons between slices yielded p>0.05). (c-d) Using 9 body models reduced the sensitivity of psSAR to positional mismatches. Nevertheless, the actual SAR was still up to 3.7-fold higher than the peak estimated using the 9 models.

Figure 4: The regions that were exposed to higher levels of local SAR than the estimated maximum due to the positional mismatch are shown for two selected positions. (a) A region of size 166 cm³ was exposed to elevated levels of local SAR at the worst-case scenario. 20 cm³ of tissue was exposed to more than twice the estimated maximum local SAR. (b) The positional mismatch caused elevated levels of SAR exposure in the right anterior part of the head in a different case.

Figure 1: Temperature vs. time for long-duration (i.e. 10s averaged) vs. short-duration B_1+RMS exposure

Figure 2: Thermal dose vs. time for long-duration (i.e. 10s averaged) vs. short-duration B_1+RMS exposure

Figure 5. Example of the first full-waveform pTx scan in vivo after following validation tests outlined in this abstract. Here, a conventional 3D GRE sequence with a CP-mode hard pulse is compared to a Universal Pulse based on a 3D SPINS trajectory. Both pulses were set to a nominal flip angle of 5° and their reported SAR values are outlined as well.

Figure 3. Comparison of prescribed pTx RF pulse (blue) and normalized measurement waveforms (magenta) of the UP applied in the center-most TR of the 3D gradient-echo sequence. The left set of columns show the magnitude plots for 4/8 channels and right columns show the corresponding phase data. Some deviations between the prescribed and measured data exist for these more complex waveforms, yet in general are still in good agreement.

Figure 1 A, The schematic diagram of experimental setup of the pulse energy method with flux loop defined in NEMA MS 8. B, The schematic diagram of experimental setup of equivalent circuit model without flux loop.

Figure 2 A, Equivalent circuit model of the empty coil in the power transmit chain. B, Equivalent circuit model of the coil loaded with subject in the power transmit chain.

Schematic of the proposed VOP algorithm. In its core it uses Lee’s algorithm (2) with a speed enhancement proposed by Kuehne et al. (4).

Comparison of the compression achieved with the original (Lee) and the enhanced algorithm for all three arrays. The left column shows the results in terms of number of VOPs, the right column the respective calculation time. For comparison: for the lowest overestimation (0.671% or 2.68%), Eichfelder’s algorithm (not in plot) concluded with 10,788 VOPs after 362 min (8-channel local array), 6,270 VOPs after 527 min (8-channel remote array), and 2,802 VOPs after 441 min (16-channel array).

Figure 1. Workflow and the experimental set-up for predicting RF heating at the contact points of commercial DBS electrodes due to RF exposure in MRI. (a) Phase 1: Offline calibration to determine the R_eq. Phase 2: Temperature prediction. (b) Position of the temperature probe around the contacts of the DBS electrode. (c) The commercial IPG device used in the experiments. (d) The extension cable with a protective boot.

Figure 4. Comparison between the simulated and measured temperatures corresponding to five different cases. RMSE of less than 0.15°C promises an accurate temperature prediction.

Figure 2. The general test set-up (GE MR750 testing). The H-field sniffer probe is within the phantom, in the center of the scanner. The probe remote unit was placed on a shelf as far away as possible from the scanner. The remote unit was connected, via a coaxial cable through a waveguide, to a standard oscilloscope positioned beside the operator console.

Figure 1. Phantom used for testing. Left: The phantom is shown filled to a depth of 7 cm for testing on the Siemens Prisma system. Right: the phantom is shown filled to a depth of 4 cm for testing on the GE MR750.

Figure 1 Distributions of the patient examination duration for different scan regions under various RF exposure levels.

Table 2 Scan time for standard protocols under Normal Operating Mode (wbSAR limited to 2W/kg) and the prolonged protocol durations when wbSAR limited to 1W/kg and 0.5W/kg.

Figure 6: T₁ MPRAGE image of Medtronic DBS electrode aligned angled to the B₀ field showing 4.7 cm wide susceptibility artifact.

Figure 1: Temperature rise for Abbott and Medtronic DBS electrode tips aligned parallel to B₀ field and scanned using GRE, T₁ MPRAGE, T₂-TSE sequences.

Number of VOPs computed for different overestimation factors λ. Blue: Lee’s generalized VOP compression method, red: this work.

10g-SAR computed over the uncompressed set of voxels and over the VOPs for the two different methods (λ=0.7) and for 1 million random RF shims. a) 10g-SAR over the VOPs versus the true SAR. The dashed black line is the identity line. b) Boxplots of the relative overestimation for the two methods. The SAR is never underestimated.

Figure 1: The proposed SAR Monitor consists of two dual directional couplers (1,2), two RMS envelope filters (3) and a microcontroller (4).

Figure 4: A block diagram of the proposed SAR monitor, that handles power limit exceedance and RF-coil errors for a standard MRI RF transmit / receive setup.

Figure 1: Anthropomorphic gel phantom filled with poly acrylic acid gel. The gel thickness was 133mm.

Figure 3: Temperature profile at the guidewire tip for the body and local coil during a high power sequence.

Figure 2: This test fixture was constructed to allow direct measurement of the impedance between an IPG’s header connector block and Can. Also shown are the IPGs modified to provide direct access to terminal blocks to allow this measurement.

Figure 1: Simplified diagram showing the RF current circulation paths that cause tissue heating.

Figure 1: Simplified model 8C1SW: 8-contact electrode with 1 shared wire (connected to the 8 contacts).

Figure 3: Estimated SAR distributions in the coronal view for: (A) simulation 1 and (B) simulation 2.

Figure 1

Figure 2

Fig. 3: Temperature rise for 150V/m average incident E-field for all four objects for both systems.

Fig. 2: Measured E-field distributions with 0dB and ±1dB isolines and the placement of the 300mm-long rod for the (a) birdcage coil and (b) linear exposure system case.

Fig.3 Worst-case local-SAR following motion for pTx pulses. (i) psSAR for worst-cases as a factor of psSAR_centre (same pulse without motion). Tissue volume exposed to higher SAR than psSAR_centre shown in orange. (ii) SAR profiles for corresponding worst-cases (bottom row) compared to their SAR_centre (top row). Colour scale shows psSAR as a factor of psSAR_centre (normalised per coil). The number of spokes in each coil’s worst-case pulse is indicated with asterisks. (iii) The positions at which these worst-case psSAR were observed. All worst-cases were observed at slice 1 or 2.

Fig.2 Motion-sensitivity of (i) peak local-SAR and (ii) whole-head SAR for RF-shim pulses designed using at the central position, and evaluated at all off-centre positions for each coil model A to F. Slice 2 shown; slices 2-5 showed similar trends. Y-axes refer to SAR at each evaluated position (left-right [L-R] and anterior-posterior [A-P] displacements), as a factor of SAR at the centre position (i.e. without motion). SAR_centre was comparable across coil models (not shown).

Figure 3: a) Power loss on electrodes vs lead length for antenna effect simulations seen in Figure 2a for copper and silicone-rubber, respectively. b) Power loss on electrodes vs conductivity at worst-case lead length. c) 1 g average local head SAR for simulations at 128 MHz (top) and 298 MHz (bottom). d) Temperature measurement for center-surround electrodes with center leads (Figure 3c). L in the legend indicates that it is a measurement on the left electrode. The numbers on the electrodes indicate the probe position referred to in table 2.

Figure 1: a) Commercially available TES/MRI setup with copper leads and 5 kΩ safety resistors. Leading the wires through the opening in the coil and using twisted pair cables restricts the lead configuration and causes stray fields compromising MRCDI experiments. b) and c) show the proposed electrodes and the lead design. Low-conductivity silicone-rubber is used for electrodes and leads thermally and electrically shielded with a glass fiber sleeving. Medical grade touch-proof MC connectors are used to connect the electrodes to copper lead wires 90 cm away from the subject's head.

Figure 1. Duke at abdomen landmark: a) centered b) shifted to his right side and c) associated unloaded coil B1+ map [uT] at hotspot level shown in Figure 4.

Figure 2. a) the variations in B1+ caused by loading [log10(loaded B1+ / unloaded B1+)] and b) the associated SAR map. Both figures are at the coronal level of the SAR hotspot in wrist for Duke in the central position. The colourbar range in a) only has been truncated by approximately 50% to magnify colour variations.