IMPROVING SNR IN SMALL TEMPERATURE CHANGE MR THERMOMETRY TO ACQUIRE SAR MAPS OF A PAIR OF ASL LABELLING COILSAaron Oliver-Taylor1, Chris Randell2, Roger J Ordidge3, David L Thomas4

1THE ADVANCED MAGNETIC RESONANCE IMAGING GROUP, DEPARTMENT OF MEDICAL PHYSICS AND BIOENGINEERING,UNIVERSITY COLLEGE LONDON, LONDON, UNITED KINGDOM2PULSETEQ LTD, UNITED KINGDOM3CENTRE FOR NEUROSCIENCE, UNIVERSITY OF MELBOURNE, MELBOURNE, AUSTRALIA4INSTITUTE OF NEUROLOGY, UNIVERSITY COLLEGE LONDON, LONDON, UNITED KINGDOM

References: [1] Mildner, T., et al.; MRM; 49 5:791–795; 2003. [2] IEC 6601-2-33, Edition 3; 2010. [3] Ishihara, Y., et al.; MRM; 34 6:814–823; 1995. [4] Oh, S., et al.; MRM; 63 1:218–223; 2010. [5] Cline, H., et al.; MRM; 51 6:1129–1137; 2004. [6] Poorter, J.D., et al.; MRM; 33 1:74–81; 1995.Acknowledgements: David Carmichael, Jeff Hand, Antoine Lutti, Oliver Josephs, Nik Weiskopf, and the UK Medical Research Council.

INTRODUCTION●Separate coil continuous arterial spin labelling (CASL)1 uses a close fitting surface coil coupled with a low power RF amplifier.●Total power is low (1-2W), however local SAR can be high if not controlled, therefore it is desirable to obtain SAR maps to ensure safe operation2.●Computational EM modelling requires complex, expensive (if commercial) software, and it can be difficult to capture the nuances/imperfections of “home-built” coils.●MR Thermal Imaging3 (MRTI) uses the temperature dependence of the proton resonance frequency (PRF) to encode temperature change into the phase of a gradient echo image.●Presented is an optimised method for producing an MRTI time series of heating from a pair of CASL labelling coils, using a 3D EPI acquisition and image processing technique for calculating high SNR SAR maps.

THEORY●If thermal conduction is negligible and the object is at thermal equilibrium when heating commences, the SAR can be calculated from the initial temperature change, as given by equation 1.●The negligible conduction only occurs for a short period of time (2 minutes [5]), and the low power used in CASL means only a small temperature rise occurs over this duration, therefore low SNR.●To increase SNR it is proposed to make regular MRTI measurements over an extended period of heating, and then fit the temperature change to a suitable fitting function.●The fitting function is required to match the temperature changes within a voxel.●Thick agar gel (therefore no convection) experiences heating due to the RF power deposition, and cooling due to thermal conduction to neighbouring voxels.●A suitable function, empirically found to match both MRTI and fibre optic thermometer measurements is given by equation 2.

RESULTS●Figures 4 and 5 show plots of the MRTI measurements averaged over each ROI (see figure 6). By temperature compensating the reference phantom phase, the corresponding MRTI measurements (orange curve) closely resembles the FO probe data (light blue curve).●Raw SAR maps (calculated directly from the MRTI data) have no useful SAR information (figures 7 and 8). In contrast the SAR maps calculated from the fitted MRTI data show realistic SAR distributions and values for the ASL Labelling Coils (figure 7-9).●The SAR maps for the multiple transmit configurations (in phase, anti phase and quadrature), match well with superpositions of the individual coil SAR maps.●The maximum local SAR was found for the In Phase transmit configuration: 8.78±0.38W/kg.

DISCUSSION AND CONCLUSION●We have developed a new method to accurately map the SAR of low power ASL coils, using MR thermometry and a model-based analysis of the time course of temperature changes.●Both FO probe and MRTI measurements show good agreement, indicating the B

0 drift

correction produces accurate temperature measurements.●SAR maps show similar spatial distributions, and the multiple transmit configurations match with superpositions of the single coil SAR maps.●Maximum local SAR measured is less than the IEC's 10W/kg local SAR limit2. No spatial-mass averaging has been performed, which would reduce the peak SAR.●The increased salt content of the gel phantom will likely mean that significantly more power was deposited than would be in muscle tissue.●The method requires a high degree of phase stability, and long scan times.●Further work is to compare results with EM simulations, improve the B

0 drift correction to

compensate for non-linear B0 drift, and to further investigate the sources of error

associated with the SAR calculation.

T fit (t )=c1 tec2 t+c3

Eq. 2: FITTING FUNCTION

Fig. 1: ASL LABELLING COILS

PLANAR PCB DESIGN 45mm ID

64mm ODTUNED AND MATCHED TO 123.2MHz.

SAR=c(T (t 2)−T (t1))

t 2−t1c = specific heat capacity (4200J/kg)

T = temperaturet1/2

= time at measurements 1/2

Eq. 1: SAR DEFINITION

METHOD: MRI●3T Siemens Tim Trio (Erlangen, Germany) Clinical MR System, transmit with birdcage body coil, receive with 12 channel head matrix coil.●2 additional low power (1W) transmitter channels built using an upcycled spectrometer (MR5000, SMIS, Guildford, UK). Triggered via optical sync pulse from the Trio.●Pair of PIN diode detunable ASL Labelling Coils (figure 1).●Agar gel phantom for heating measurements, 3 liquid reference phantoms for first-order B

0 drift correction6 (figure 2).

●Point measurement of the temperature of each phantom made using a fibre optic (FO) thermometer (ProSens with PSP-62 Modules and OTP-A Sensors, OpSens, Quebec, Canada).●3D GE EPI volumes acquired every 33.84s: FOV = 210 × 210mm, 2.5mm slice thickness, TE=32.86ms, TR=80ms, and GRAPPA 2× acceleration, TA=3840ms.●100 acquisitions were made, with 30s of RF power applied between.●RF applied 100kHz above the imaging frequency to avoid saturation of spins●Five transmit configurations were applied, given in table 1, two sets of measurements per configuration.●Prior to each measurement session the same EPI sequence was run for a minimum of 2.5 hours to “preheat” the passive shims, improving B

0 stability during MRTI

measurements, and to allow the phantoms to reach thermal equilibrium.●Images were processed in Matlab (The Mathworks Inc., Natick, MA) as described by figure 3.

GEL PHANTOM12CM Ø, 20CM LONGDISTILLED WATER2% AGAR1.6% CuSO

4

0.9% NaCl

3x REFERENCE PHANTOMSINSULATED

3.2CM Ø, 25CM LONG1.6% CuSO

4

Fig. 2: PHANTOM

Amplitude PhaseA

1A

2φ

1φ

2

Right Coil 100% 0%(detuned)

0° N/A

Left Coil 0%(detuned)

100% N/A 0°

In-Phase 100% 100% 0° 0°Quadrature 100% 100% 0° 90°Anti-Phase 100% 100% 180° 180°

Table 1: TRANSMITCONFIGURATIONS

PHASE, ϕPHASE

INCREMENT, Δϕ

Δϕ=arg (e iϕn

e iϕ1)

COMPLEX DIVIDEBY FIRST

MEASUREMENT

TEMPERATURE CORRECT

REFERENCE PHANTOM PHASE

REFERENCE PHASE

SEGMENT Δ INTOϕPHANTOM AND

REFERENCE PHASE

B0 DRIFT

CORRECTION:SUBTRACT

CORRECTION MAP FROM

PHANTOM Δ ϕ

2D FIRST ORDER

POLYNOMIAL FIT

CORRECTED Δϕ

T=Δϕ

αγ B0TE

COMPUTE TEMPERATURE MAPS

α = -0.0097ppm

TEMPERATURE MAP

FIT VOXELS TOEQUATION 2

COMPUTESAR FROM

INITIALTEMPERATURE

INCREASE

SAR MAP Fig. 3: IMAGE ANALYSIS

CORRECTION MAP

EPI VOLUMETIME SERIES

FIBRE OPTICTHERMOMETER

DATA:SMOOTH AND

INTERPOLATE TOMR ACQUISITION

TIMES

Fig 9.a: LEFT COIL SAR Fig 9.b: IN PHASE SAR Fig. 9.c: QUADRATURE SAR

RAW SAR FITTED SAR RAW SAR FITTED SARFig. 7: RIGHT COIL SAR Fig. 8: ANTI PHASE SAR

Fig. 4: RIGHT COILTEMPERATURE

Fig. 5: ANTI PHASETEMPERATURE

Fig. 6: ANTI PHASEROI POSITIONS

PROBE ROI: 9x9 VOXEL ROI AROUND THE PROBE TIP.

COIL ROI: 9x9 VOXEL ROI POSITIONED CLOSE TO THE TRANSMITTING COIL.

SIMILAR ROIs WERE USED IN ALL OTHER CONFIGURATIONS