11 April, 2003 PMRIL Autotuning Interventional Coils for Imaging Ross Venook, Greig Scott, Garry...

11 April, 2003PMRIL

Autotuning Interventional Coils for Imaging

Ross Venook, Greig Scott,

Garry Gold & Bob Hu

11 April, 2003PMRIL

Motivation for Automatic Tuning: Clinical Realities

• Interventional applications = uncontrolled variables– Flexible coils are clinically desirable

– Stuff is inhomogeneous and uncooperative inside the body (even an unconscious subject has moving things under the surface)

• Specific interventional applications– ‘deployable’ RF coil

– Range of motion studies

11 April, 2003PMRIL

Motivation for Automatic Tuning: Technical Consequences

• RF coils are basically inductive loops with a tuning capacitance– Changing the shape or size of the loop changes the

inductance (and hence its tuned peak), but manual adjustment of capacitors is slow

• Closer coupling between the coil and the load means increased coil dynamics

• SNR depends on coil tuning (matching)

Goal: Create an automatic tuning device to quickly and easily optimize coil SNR

11 April, 2003PMRIL

Agenda

• Motivation

• Background– Autotuning method– Electronics development– Old results

• Autotuning with Signa

• Theoretical SNR analysis

• Future work

11 April, 2003PMRIL

Background I: ‘Autotuning’

• Autotuning is the process by which a tuned coil’s center frequency is adjusted without manual effort (push-of-a-button)

• Many methods/topologies exist which can accomplish the task of tuning (and matching) a resonant circuit– Probe topology important

– Tuning topology important

11 April, 2003PMRIL

Background II: Autotuning Method

• Tuned elements have a complex impedance• ‘Resonance frequency’ is defined by zero

imaginary impedance

50 55 60 65 70 750

10

20

30

40

50

60

Frequency [MHz]

Res

ista

nce

[Ohm

s]

50 55 60 65 70 75-30

-20

-10

0

10

20

30

Frequency [MHz]

Rea

ctan

ce [O

hms]

11 April, 2003PMRIL

At 63.9MHz

0

10

20

30

40

50

60R

esis

tanc

e [O

hms]

-20

-10

0

10

20

30

Rea

cta

nce

[Ohm

s]

50 55 60 65 70 75

10

20

30

40

50

60

Frequency [MHz]

Res

ista

nce

[Ohm

s]

50 55 60 65 70 75

-20

-10

0

10

20

30

Frequency [MHz]

Rea

cta

nce

[Ohm

s]

10

20

30

40

50

60

Res

ista

nce

[Ohm

s]

-20

-10

0

10

20

30

Rea

cta

nce

[Ohm

s]

0

0

11 April, 2003PMRIL

Electronics:Varactor-tunable RF Coil

75nH

22 or 68pFVaractor

Q spoil SignalPort<360nH

C

9 V

20K 20K150pF

10K

C

DC Tuning Bias

Signal

Flex CoilDC Tuning Bias

11 April, 2003PMRIL

Electronics:Reactance Detection

• Phase comparator outputs a DC voltage that is a function of probe reactance

-600-400-200

0200400600

55 57 59 61 63 65 67 69

Frequency (MHz)

DC

out

put (

mV

)

11 April, 2003PMRIL

Electronics: Microcontroller

• Atmel 90S8515 microcontroller operates the state machine via SPI (serial peripheral interface)

FrequencySynthesizer

Micro-Controller

PhaseDetector

Scanner Pre-amp

Tune/ReceiveSwitch

VaractorTuningVoltage

RF Coil

11 April, 2003PMRIL

Retuned

Retuned

Frequency (MHz)

Res

ista

nce

(Ω)

61 62 63 64 65 66 67

10

20

30

40

50

60

DetunedDetuned

Punchline

11 April, 2003PMRIL

Agenda• Motivation• Background• Autotuning with Signa

– Initial experiments: RFI, and other problems– Improved autotuner– Initial experiments (for real, this time)– SNR increase observed (!)

• Theoretical SNR analysis• Future work

11 April, 2003PMRIL

Initial Images (Problems)

• RF interference artifact caused by switching power supply

• Common-mode

transients affect

microcontroller,

TR switch fails

while imaging

11 April, 2003PMRIL

Solution• Replace switching supply and 3V Li battery (very

cool, non-ferrous)…

• …with simple voltage regulators and 9V alkalines (not cool, very ferrous) and….

11 April, 2003PMRIL

…goodness results

• Reliable images with autotuner (Greig at Leipzig)

High peak SNR (>300) ‘No’ RFI, some PE ghosting

Windowed-down Windowed-up

(SPGR, TE/TR = 7.2ms/34ms, 4mm slice, 12x12cm² FOV, 30° flip)

11 April, 2003PMRIL

Experimental Setup

Varactor-tunedCoil

Phantom

AutotuningElectronics

cable

GE Signa1.5 T

Status LEDs(not blinking)

11 April, 2003PMRIL

Experiment #1The Pepsi Challenge

• Varactor-tuned vs. passively-tuned images Varactor-diode

Passive variablecapacitor

Both images have nominal SNR=326

11 April, 2003PMRIL

Experiment #2:Tune-Detune-Retune

Autotune Detune (deform coil) Re-autotune

11 April, 2003PMRIL

SNR Profiles

Noise Box Image Sample

SNR

11 April, 2003PMRIL

Typical ResultSN

R

Autotune Detune (deform coil) Re-autotune

11 April, 2003PMRIL

Agenda

• Motivation• Background• Autotuning with Signa• Theoretical SNR analysis

– Noise Figure and noise circles– Coil -> preamplifier– Coil -> txn line -> preamplifier– An interesting result

• Future work

11 April, 2003PMRIL

Noise Figure

• All practical devices have NF>0dB (F>1)• Convenient and sensible metric• Noise Figure is in dB, Noise Factor is not

10log( )

IN

OUT

SNRF

SNR

NF F

11 April, 2003PMRIL

Why We Have Preamplifiers

• Friis Equation

• Preamplifier NF dominates system NF (for moderate G1)

– Worry about preamp NF

F1, G1 F2, G2 FN, GN…SNRIN SNROUT

Fi = Noise Factor of ith stage Gi = Gain of ith stage

321 1

1 1 21

1 11IN NN

OUT nn

SNR F FFF F

SNR G G G G

11 April, 2003PMRIL

Preamplifier NF• Depends on impedance match, Ropt set by device

R/Ropt

Noi

se F

igur

e [d

B]

Fmin = 1.05, 1.07, 1, 1.2, 1.5

1

2

3

4

5

6

7

0.1 1 10

11 April, 2003PMRIL

Preamplifier NF• Actually, NF is a surface on the complex-Z plane

Im[Z]

Re[Z]

NF

(dB

)

11 April, 2003PMRIL

Noise Circles

• Preamplifier NF level sets are circles in the complex impedance plane

• Minimum NF occurs at 50 + j0 Ω for this example

Im[Z

] (Ω

)

Re[Z] (Ω)0 1000

0

-500

500

Preamplifier NF Contours [dB] on Z-plane

11 April, 2003PMRIL

Autotuning Trajectory

Im[Z

] (Ω

)

Re[Z] (Ω)0 1000

0

-500

500

Preamplifier NF Contours [dB] on Z-plane

•••

•

• •

•• •

•••

11 April, 2003PMRIL

Lossy Transmission Lines

• Coaxial transmission lines connect the coil to the tuner, and the tuner to the scanner– Impedance transformation changes Z at preamp

– Resistive loss adds noise

• Interventional devices require small-diameter coax– Greater loss (dB/m)

– Imperfect impedance transformation

– Together, these warp the system noise circles and trajectories for a given coil

11 April, 2003PMRIL

Lossy Transmission Lines (cont…)

• Evaluate impedance ‘mismatch’ by measuring reflected power from a transition– Perfect match has no reflections

• Statement: ‘lossier cable can improve your match because it lowers the reflected power’

• Hmmm….• Counter: ‘what about SNR? Isn’t the baby being

thrown out with the bathwater??’

11 April, 2003PMRIL

Interesting Result

• Indeed, lossy transmission line will always have worse SNR at its output than at its input

• But, if we have a preamplifier waiting for us on the other end, perhaps the impedance transformation can improve system SNR

• Conclusion:– It’s something to watch out for

Noi

se F

igur

e [d

B]

1

2

3

4

5

6

7

0.1 1 10

R/Ropt

11 April, 2003PMRIL

Future Work

• Loose ends– Controlled analysis of tuning vs. SNR with data

• Theoretical– To tune, or not to tune (and how to decide)

• Clinical– Scan cadaver shoulders

• Practical– 0.5T/21MHz version– Test speed limits

• New directions– Automatic matching (perhaps a successful topology to borrow

from CW-EPR)