An 8-channel RF Transceive Coil For Parallel Imaging at 7T ... · • CAD model imported in MWS...

An 8-ch RF transceive coil for MRI at 7T/298MHz X. Hanus CEA-DSM-Irfu-SACM 1

An 8-channel RF Transceive CoilFor Parallel Imaging at 7T/298MHz

X. HanusCEA – DSM – Irfu – SACM

CST European User Group Meeting 17-19 May 2011, Munich, Germany


Contents

• NeuroSpin: a high-field MRI facility at CEA Saclay

• Iseult: a franco/german R&D program for high field MRI

• RF coils design with numerical codes

• MWS-TD comparison for an 8-channel RF prototype coil with two anatomical models, Ella (Virtual Family) and Hum (Home-made Unstructured Model based on Ella tissues):

• mesh & transient simulation parameters• tune/match circuit in Design Studio• illustration with a reference birdcage mode (rotating B1

+ field)


NeuroSpin: a neuro-imaging MR facility at CEA Sacla y

• Project initiated in 2003 from the CEA Life Sciences Division (DSV) to achieve a unique platform for high field MR imaging

• Internal collaboration with CEA Physical Sciences Division (DSM)

• Franco/german 'Iseult' collaboration (CEA/Siemens/partners)

• Building delivered in 2006 (Vasconi).

• Actually: 3T & 7T scanners for human (Siemens),

17T scanners for small animals (Bruker),

11.7T/90 cm human MRI scanner (Iseult, 2013)


NeuroSpin: a neuro-imaging MR facility at CEA Sacla y

• Building delivered in 2006 (Vasconi).


Iseult/Inumac partnership goals

• Iseult/Inumac: a franco-german CEA-Siemens R&D program for human MR imaging at Ultra High Field (11.7T/500 MHz MRI scanner, contrast agents, pulse sequence designs, etc.)

• Contract Iseult / CEA-DSM-Irfu to design and build a large bore 11.7T/90 cm magnet, cryogenic system and RF coils (SACM)

• A challenging transmit RF coil for parallel imaging at 500 MHz

• Maintain B1+ homogeneity over the brain despite wavelength effects (as

frequency increases, RF wavelength is shortened (λ0=60 cm at f=500 MHz) and even more in the brain (ε=50) → λbrain=8.5 cm

• Limit RF dissipations and heating in the inhomogeneous and dissipative dielectric human tissues (SAR, hot spots)

• A test bed with a validated design at 300 MHz (7T)


Iseult/Inumac partnership organization

Oséo BMBF

Consortium français

Univ de Fribourg

Siemens MS

Bruker Biospin MRI

Franco-German Co-Operation Agreement

mandat

Acc Collab Acc Collab

Acc Collab

Consortium allemand

25 M€CEA

Alstom MSA

Guerbet

Conv.d’Aide

Acc Collab

Luvata

Acc Collab


RF coils design with 3D full-wave solvers

• Various coil designs simulated with MWS-TD since 2004:• a 500 MHz TEM/birdcage volume coil with 32 rods, • 300MHz array coils based on striplines superposition

• Curent designs: 8 channels, superposed striplines with Balun feed, and pre-matching system⇒ FEM solver with conformal (tetrahedron) meshing to design and optimize the strips assembling⇒ MWS-TD validation with accurate anatomical voxel models (HUGO, Virtual family) or home-made models• CAD model imported in MWS (.sat/.hfss) • Setup for transient calculations (Boolean operations, coaxial

connectors (50 Ω) to apply waveguide excitations, additional inactive volumes for local meshing)

• tune/match circuit in Design Studio → maps (E,H,J) x 8 channels• basic birdcage mode (combine) → B1, SAR


The FEM model and results

B1+ (a= 6.8)

SAR_CP (a= 6.8)S_I,I S_I,2S_I,1


The FIT model (import + coaxial WG ports) (1)

Different views of FIT model


The FIT model: Coil without shielding (2)


The FIT model: Balun and coaxial WG port (3)


The FIT model: radiating dipole and capacitor (4 )


The FIT model: WG port to stripline transition (5 )


Design model and Cartesian grid

2D mesh view of dipole-1 and dipole-2 in mid-axial plane (left), and at port plane (right)

Mesh grid details for dipole-1 in axial planes (20 Mcells, local Dx,y = 0.7mm)

• Align all fixpoints to mesh grid is not possible (< 0.3mm)• Meshing is not identical for dipoles (diff. angle positions in the grid)• Some mesh cells may contain more than 2 materials (PBA)


Inactive volumes to improve local meshing


The anatomical models 'Ella' and 'Hum' (1)

'Ella' from Virtual Family (1mm resolution ) 'Hum' a 9-tissue Home-made Unstructured Model based on Ella dataset


The anatomical models 'Ella' and 'Hum' (2)

RF coil loaded with 'Ella_VF' (top) and 'Hum' (bottom)


Meshing parameters

Same global & local meshing parameters but Hum has finally more meshcells (fixpoints)


Mesh grid for Ella_VF-1mm


Mesh grid for Hum


Simulation characteristics

• Dell 690 XP 64 bit (4x Xeon 5140) + 2xGPU (FX Quadro 5600)• Mesh parameters:

• Global: l/λ= 30/10/30 + FPBA + Equilibrate mesh=1.2• Local: Dx=Dy=0.7 mm, Dz= 1mm (inactive volumes)• Local: prority=1 (coaxial connectors, critical elements)

• → 18.595.200 cells (Ella_VF) and 33.154.615 cells (Hum)• Transient parameters:

• WG excitations [200-400 MHz] on coaxial connectors • accuracy limit = -50 dB

• Background=PEC and BC='electric' all except zmin ('open')• Time duration ∼ 15H to 30H / port• Tuning/matching of the dipoles with a DS optimized circuit


Which accuracy limit to use ? (1)

ProtoV3b_(A) coil loaded with a sphere Inactive volumes for local meshing Global meshing Local meshing

• Coil elements – all lossy materials except coaxial connectors,• Background = PEC ; BC = (Et=0) @ all boundaries,

• Channel-2 single excitation (τP= 35.55 ns, ∆τ=1.02 ps [200-400 MHz]),

• Stop accuracy limit= -40/-50/-60/-80 db


Which accuracy limit to use ? (2)

S2,2 & 1D-probes at sphere center vs accuracy limits

Energy: stop @ -40/-50/-60/-80 db 1D-probe H_abs @ [0;0;20] 1D-probe Habs @ [0;0;20] zoom

Reflected S2,2 parameter (db) 1D-probe E_abs @ [0;0;20] 1D-probe E_abs @ [0;0;20] zoom


Which accuracy limit (3)

P0= 0.5 W (rms) -40db -50 db -60db -80db

SA

R

Ca

lcu

lati

on

Re

sult

s

Pacc. (W-rms) 0.244 25 0.239 83 0.239 53 0.239 68

Pabs. (W-rms) 0.189 32 0.190 37 0.189 80 0.189 50

Ptissue (W-rms) 0.038 44 0.038 95 0.038 90 0.038 90

SAR_10g (W/kg) 0.015 03 0.015 23 0.015 21 0.015 21

Max SAR (W/kg) 0.095 68 0.094 78 0.094 21 0.093 94

Q_

Loss

Ca

lcu

lati

on

Re

sult

s

W (J) 4.680 x 10-8 4.697 x 10-8 4.679 x 10-8 4.669 x 10-8

Pmet. (W-rms) 0.081 54 0.081 82 0.081 50 0.081 32

Pdiel. (W-rms) 0.180 98 0.182 06 0.181 51 0.181 23

Ptot. (W-rms) 0.262 52 0.263 88 0.263 02 0.262 55

Table 1: CHANNEL-2 single excitation @ f= 298 MHz for E=-40db/-50db/-60db/-80db

Table: Losses calculations results vs accuracy limits

• Accuracy limit = -50 db • Power balance consistency (effective input power at ports (S-matrix) Vs

total losses in the structure ⇒ over-estimation of metallic losses ?)


Boundary condition at z-limits: 'electric' vs 'open' (1)

BC: 'electric' @ all boundaries BC: 'open' @ xmin,xmax,ymin,ymax,zmax and 'electric' @ zmin

Scattering SI,2 parameters for 'electric condition Scattering SI,2 parameters for 'open' condition

S-parameters for port-2 for BC= 'electric' (left) and 'open' (right), background= PEC


Boundary condition at z-limits: 'electric' vs 'open' (2)

Energy decay at port-2 Scattering S i,2 (db)

1D-H probe at sphere center 1D-E probe at sphere center

Energy decay, S-parameters at port-2, 1D-H & E probes at center of coil/sphere


Effect of background material: PEC vs vacuum

Background= PEC BC = ‘electric’ S-parameters at port-1

Background= vacuum BC = ‘open’ S-parameters at port-3

S-parameters at port-1 & port-3 for background=PEC & BC= ‘electric’ (solid lines) and background= vacuum & BC = ‘open’ (dashed lines) with Ella_VF)

• RF coil has lateral shielding ⇒ ‘open’ @ zmin only

An 8-ch RF transmit coil for MRI at 7T/298MHz (X.H.,M.L.) CEA-DSM-Irfu-SACM 28

Improve the frequencies consistency

• Some improvement in the design handling:

• Larger coaxial (0.9/3 → 1.5/5), inactive cylinders better adjusted, lines/λ…

• Gives better consistency but there is no unique parameter

• ⇒ tuning of the dipoles frequencies can be made in Design Studio


Tuning/matching circuit optimized in Design Studio

• A parameterized (Z,C) circuit in Design Studio (ELT & Capa) to tune/match the individual channels

• Parameters are optimized in minimizing the reflected S-parameters magnitudes for each channel (single excitation)

• 3D-field quantities in MWS are extracted with the optimized tuned/matched circuits in DS

The tune/match Transmission Line circuit (Z,C) Ella_VF Hum


Tuning/matching: reflected S-parameters (1)

Effect of the tuning/matching on the reflected S-parameters for Ella_VF (top) and Hum (bottom)


Tuning/matching: reflected S-parameters (2)

Tuned/matched S-parameters phase and impedances (Smith) for Ella_VF


Tuning/matching : transmission to neighbours (Ella_ VF)original

After

tuning/matching

CHANNEL-1 (db, phase) CHANNEL-2 (db, phase)

Transmitted S_parameters before (top) and after tuning/matching (bottom) for CH-1 and CH-2


Tuned/matched CP: birdcage rotating field B 1+ (1)

• Basic birdcage rotating mode (a=1, phi= angle position) in D.S. • Normalization to produce a left handed 2D mean value |H1

+| = 4.68 A/m in a target square [±20,±20,20]) (90°1ms square pulse)• Combine a birdcage mode with amplitude a in a S-parameters task in

DS, calculate left handed polarization H1+| in [x,y] and evaluate the 2D

mean-value in target square

Normalized left handed |H1+| in axial plane [z=20] for 'Ella_VF' (left) and 'Hum' (right)

CP1_neg (a=8.8) CP1_neg (a=6.8)


Tuned/matched CP: birdcage rotating field B 1+ (2)

Normalized 90°1ms |H 1+| in mid-axial and sagittal planes with 'Ella_VF' (top) and 'Hum' (bottom)


Tuned/matched CP: PLD and SAR_10g (3)

PLD (left) and SAR_10g (right) for 'Ella_VF' (top) and 'Hum' (bottom) in axial plane [z=0]


Tuned/matched CP: PLD and SAR_10g (4)

PLD (left) and SAR_10g (right) for 'Ella_VF' (top) and 'Hum' (bottom) in sagittal plane


Conclusion

• Full CAD model simulation for RF coil

• Dissipative elements

• Superposition of thin strips elements

• Integrated physical pre-matching system

• Above all, no element aligned with Cartesian grids (PBA)

• ⇒ dispersion of the dipoles resonant frequencies

• ⇒ tuning/matching circuit is essential

• Small mesh steps, small time steps, low decay, multi-channel

• ⇒ prohibitive calculation time

• ⇒ GPU option is required

• Two kind of solver are necessary (FEM, FIT)

• Convergence of the simulations results FEM/FIT (sphere, Hum)

• High res. anatomical model only needed for local investigation

An 8-channel RF Transceive Coil For Parallel Imaging at 7T ... · • CAD model imported in MWS...

Documents

Transcript of An 8-channel RF Transceive Coil For Parallel Imaging at 7T ... · • CAD model imported in MWS...