1

Thermal Dosimetry Issues

In Thermal Therapy

R. Jason Stafford, PhDDepartment of Imaging Physics

AAPM Ultrasound Symposium 2011 (Vancouver, BC)

• Facilitate more optimized treatment

– planning

– targeting/localizing

– monitoring/control

– verification

• Imaging information synergistic with

integration of model based simulation

• Endgame

– increase safety + efficacy

– facilitate minimally invasive approaches

previously not considered possible/safe

Role of image guidance in thermal therapy

— Treatment prescription

Prescribed sonication point

Thermal dose (point)

— Thermal dose (total)

HIFU Ablation in Rabbit Paraspinal

Muscle @1.5T

VX-2

Spinal cord

Hazle JD, Stafford RJ, Price RE ,JMRI 15 (2): 185-94., 2002.

Thermal therapy energy sources

• Cryotherapy

• Radiofrequency

• Microwave

• Laser

• Ultrasound

Multi-element U/S applicators

High intensity focused ultrasound

(FUS or HIFU)

Focused ultrasound beam

Challenges for real-time monitoring

4

mm

3 mm

Water cooled interstitial or transurethral ultrasound applicator

30

mm

25

mmDt ~ 10 sec Dt ~ 10 minutes

Externally focused high-intensity ultrasound kernalHIFU Interstitial/Intracavitary

Faster delivery

High resolution

Small Volume

Slower delivery

Lower resolution

Larger Volume

2

Modalities for image-guided thermal therapy

• US

• CT

• MRI

Non-invasive

Non-ionizing

Arbitrary oblique plane orientation

Near real-time acquisition speeds

Multiple contrast mechanisms for

- anatomy

- function

- metabolism

- temperature …Example: MRgLITT in canine brain

Modalities for image-guided thermal therapy

T2 Pre T2 Treat T2 Post T2 Post

A pplicatorA pplicator

— Isodose: t43 = 50 m in.

Tissue Pathology

T1 Pre T1 Treat T1+C Post T1+C Post Path & D oseT1 Pre T1 Treat T1+C Post T1+C Post Path & D ose

• US

• CT

• MRI

Multi-element U/S applicator therapy

MRgFUS

MR Temperature Imaging (MRTI)

• Diffusion

• T1-relaxation

• PRF Shift

• Proton resonance frequency (PRF) of

water shifts linearly with temperature

• Sensitivity: -0.01 ppm/°C (water)

• Advantages

– Reasonable temperature sensitivity

– Relatively independent of tissue type

– Fast, gradient echo based acquisitions

• Disadvantages

– Less sensitive at low field strengths

– Lipid is insensitive to temperature

– Sensitive to background field changes

• Motion, susceptibility, etc

12s HIFU sonication

Modern Review: Rieke, V. & Butts Pauly, K, JMRI, 27:376-90 (2008);

Tissue Type

(Canine)

Temp. Range

(ºC)

Temp. Sensitivity

(ppm/ºC)

Brain 25-59 -0.0102 + 0.0005

Prostate 32-59 -0.0099 + 0.0004

Kidney 35-54 -0.0103 + 0.0006

Liver 35-51 -0.0098 + 0.0002

Bone (femur) 17-57 -0.0109 + 0.0002

Am

plitu

de (

A.U

.)

Frequency (ppm)

0 1.2 2.4-1.2-2.4

aliased lipid (-CH2)

water

(-OH)

Predicting outcomes using temperature history

• Thermal damage is cumulative effect

– Isotherm characterization of bioeffects limited

• Damage as function of exposure can be modeled as an Arrhenius rate process (W)

R = Universal Gas Constant

A = Frequency Factor (3.1 x 1098 s-1)

Ea = Activation Energy (6.3 x 105 J)

• Cumulative Equivalent minutes @ 43°C (CEM43)– Empirically derived from isoeffects observed in low

temperature hyperthermia work:

laser fiber

Normal Canine Brain5ºC

30ºC

DT

( )

0

aEt

RTA e d

W (Henriques FC, Arch Pathol, 1947; 43: p. 489.)

isodose models

W > 1

T > 57ºC

CEM43 > 240 min

( 43 )

43

0

0.25 43C EM ( ) R , w ith R =

0.50 43

n

n tnT

n

t n

T Ct t

T C

D

D

Sapareto SA, Dewey WC Int. J. Rad. Onc. Bio. Phys. 10: 787-800, 1984.

3

Arrhenius rates and a connection between A and E

Dt T for W=1

1s 60.1°C

5s 57.2°C

30s 54.1°C

240m 43.8°C(from regression)

85 points plotted from literature reports of Arrhenius fits

A variety of tissues, cells, biological macromolecules and endpoints

and heating rates over the range of 43°C-80°C are represented

Doubling Dt results in the familiar ~1°C change for equivalent

damage of CEM43

ln(A*Dt)=ln(W)-E/RT

E (Joules)

ln(A

*D

t)

Wright NT, J Biomech Eng. 2003 Apr;125(2):300-4

25 W

50 W

75 W

100 W

125 W

30ºC

0ºC

DT

T2-W

(PRF-MRTI)

T2-W FSE

(post)

T2W-FSE

(t43>240min)

Pathology/H&E

(Cogaulation,Edema, t43>240min)

Imaging versus Histology

15s FUS exposure in vivo(skeletal muscle)

1cm

Regression and Bland-Altman Analysis

Bias: 54.0+299.4 mm3

ca

nin

e p

ros

tate

ca

nin

e b

rain Bias: 7.35+20.1 mm2

Dice Similarity Coefficient

(Yung, JP, et al, Med Phys. 2010 Oct;37(10):5313-21)

Arrhenius Dose (W>1) ()

Outer rim of T1+C enhancement ()

Subject DSC

1 0.92

2 0.91

3 0.91

4 0.93

mean+s 0.92+0.01

Dice similarity coefficient (DSC):

2# in tersecting

average

X YDSC

X Y

Arrh

eniu

s Dam

age (W

)

DT

(°

C)

4

ΔT

1-W

Am

pli

tud

e

(%)

Te

mp

era

ture

(°C

)

FCSI for Multiparametric Imaging Guidance

70°C

37°C

50 Hz

0 Hz

2.5%

0 %

ΔR

2*

(H

z)

Isotherm lines for damage from the Arrhenius

dose rate model correlate well with areas where

inflection is seen in the slope of R2*(T).

Calculated Dice Similarity Coefficient (DSC) = 0.84

fiber

Taylor, BA, Stafford, RJ, iMRI Symposium 2010 (submitted to NMR in Biomed)

Arrh

en

ius

Do

se

{W/(1

+W

)}

T (C)

R2*

(Hz

)

DR2* Derived

Arrhenius

Logit dose analysis of ex vivo changes in tissue R2*

W(LD50) = 0.77+ 0.40 W(LD50) = 1.10 + 0.32 W(LD50) = 0.85+ 0.49

W(LD50) = 0.97+ 0.30 W(LD50) = 1.10+ 0.25 W(LD50) = 1.00 + 0.14

Probit analysis of thermal dose: McDannold NJ, et al, MRM 2004; Above results: Taylor BA, et al, JMRI 2010.

Arrhenius rate processes for alternative isoeffects

W>1

Vascular stasis

(Brown 1992)

W > 1 (skin)

(Pearce SPIE 2009)

W > 1 (skin)

(Henriques 1943)

DT> 54°C

(Isotherm for 30sec)

T1+C (sub)

(15 min heating)

W>1

N-K pump

(Grinberg 2001)

CEM43 > 240 min

(Sapareto-Dewey 1984)

FD > 0.5

(Pearce, SPIE 2009)

Mu

lti-

ele

me

nt

IUS

he

ati

ng

(15

min

@ 7

W)

Arrhenius rate processes for alternative isoeffects

CEM43 > 240 min

(Sapareto-Dewey 1984)

FD > 0.5

(Pearce, SPIE 2009)

W > 1 (skin)

(Henriques 1943)

DT> 54°C

(Isotherm for 30sec)

T1+C (sub)

(15 min heating)

W > 1 (skin)

(Pearce SPIE 2009)

W>1

Cyt C

(Liggins 1999)

W>1

ATP synth e

(Wang 1993)

Mu

lti-

ele

me

nt

IUS

he

ati

ng

(15

min

@ 7

W)

5

Arrhenius rate processes for alternative isoeffects

CEM43 > 240 min

(Sapareto-Dewey 1984)

W>1

ATP synth e

(Wang 1993)

FD > 0.5

(Pearce, SPIE 2009)

W > 1

(Pearce, SPIE 2009)

W > 1 (skin)

(Henriques 1943)

DT> 54°C

(Isotherm for 30sec)

T1+C (sub)

(15 min heating)

W>1

HSP70 Expression

(Beckham 2004)

Mu

lti-

ele

me

nt

IUS

he

ati

ng

(15

min

@ 7

W)

rf

Gs

Gf

GΦ

16-echo MFGRE for fast CSI

Fast Chemical Shift Imaging (FCSI) for MRTI

-25

-20

-15

-10

-5

0

5

10

15

20

0 100 200 300 400 500 600 700 800

CSI (Difference)

CSI (Water Only)

CPD

Flouroptic Probe

Time (sec)

Tem

pera

ture

Chan

ge (

°C)

ΔT(δW-L)

ΔT(δW)

ΔT(Δ φ)Fluoroptic Probe

Susceptibility induced error

corrected using lipid reference

1

( ) ( )n n n

Ni d TE

n

n

FID t C e w t

0

1

( )( )

( )1

P

n

n

n

Q

n

n

n

zB z

H zA z

z

Im ln( )

R e ln( )

( )

( )

n

n

n

n

n

n

n n

t

dt

BC

A

D

D

water

CSI-MRTI in a Canine Femur

DT

30ºC

5ºC

Temperature Probe

for verification

Am

plit

ude (

A.U

.)

Frequency (ppm)0 1.2 2.4-1.2-2.4

MR Spectrum (1 voxel from image)

lipid peaks

water

Taylor BA, et al., Medical Physics. 2008;35(2):793-803.; Taylor BA, et al., Medical Physics. 2009;36(3): 753-764.

PRF-MRTI of FUS in breast

Multi-planar, multi-shot EPI MRTI

facilitated real-time MRTI with

high spatiotemporal resolution,

high SNR and lipid suppression

Fast CSI for MRTI

Water Water and Lipid

3-Peak Lipid

3-Peak Lipid

3-Peak Lipid

6

Laser

Fluoroptic

Probe

Adipose

Tissue

T1-W MRTI for Adipose Tissue

TSC = -0.56 +/- 0.05 %/C

Tem

pera

ture

(C

)

T1-

W S

igna

l (R

2*

corr

ect

ed)

Temperature (C)

Adipose Temperature Sensitivity Calibration

Temperature (C)

Adip

ose

R2

* (H

z)

Arrh

enius Dose (W

)

breakpoint at W=1

Gel

Hynynen K, McDannold N, Mulkern RV, Jolesz FA., MRM 2000 Jun;43(6):901-4.

Taylor BA, et al, NMR Biomed (2011)

Fast CSI of FUS in canine prostate

Water PRF

0

5

10

15

20

25

30

W-F T2*

Challenges for PRF MRTI in the Body

• Motion distorts both anatomy & magnetic field

– Need to correct both for integrated damage estimate

– Breath holds can provide reasonable results

• Background B0 correction

– Gating, navigators, multi-baseline techniques

– Internal reference techniques (i.e., lipid)

– Referenceless techniques (Rieke V, et al 2004)

– Real-time modeling and simulation with MR feedback

magnitude temperature

(uncorrected)

temperature

(corrected)

mean magnitudeArrhenius Dose

(uncorrected)

Arrhenius Dose

(corrected)

Re

al-

tim

e

(in

sta

nta

ne

ou

s)

Re

al-

tim

e

(cu

mu

lati

ve

)

77ºC

42ºC

Below: MR-guided laser

ablation in liver with

background phase-correction

Challenges

Background field drift

Intra/inter scan motion

Through plane motion

Applicator susceptibility

Variable SNR (tissue/time)

Coupling MRTI feedback to models

Pennes Bioheat Equation useful for modeling therapy:

HEAT SOURCES (laser, ultrasound, RF, etc) :

P = Power absorbed (W m-3)

HEAT CONDUCTION (diffusion):

T = temperature (Kelvin) ρ = density (kg/m3)

c = specific heat of material (J kg-1 ºC-1) k = thermal conductivity of tissue (W m-1 ºC-1)

HEAT CONVECTION (effects of perfusion):

b = blood density (kg m-3) Cb = specific heat of blood (J kg-1 ºC-1)

V = volume flow rate per unit volume (s-1) k = dimensionless convection scale factor

( , )

[( ( , ))] 1 ( , )ii b b a i

T q tc k T q t C V T T P q t

t k

Pennes HH, J Appl Physiol. 1: 93–122 (1948)

heat diffusion heat convection heat source

7

Dynamic Data Driven, Model Constrained MRTI

MRTI

(°C)

Uncertainty

(°C)

Simulated

Data Loss

Kalman Predicted MRTI( 95% CI for 57°C Isotherm)

Measured MRTI(°C)

MRTI = Dynamic Input

Algorithm compares

uncertainty in input

verses uncertainty in

model and provides an

optimal estimate

Can help compensate

for noisy, missing or

corrupt data

Can help in optimal

interpolation or

retrospective filtering

Computation intensive,

requires highly

parallelized, high

performance computing

to perform in real-time

(work in progress)

David Fuentes, PhD

MDACC

Imaging Physics

• John Hazle, PhD

• Edward F. Jackson, PhD

• R. Jason Stafford, PhD

• Luc Bidaut, PhD

• Ken-Pin Hwang, PhD

• Jim Bankson, PhD

• Andrew Elliot, PhD

• David Fuentes, PhD

• Marites Melancon, PhD

• Brian Taylor

• Joshua Yung

• Krista McAlee, RT(R)MR

• Brandy Reed, RT(R)MR

Experimental Diagnostic Imaging

• Chun Li, PhD

• Marites Melancon, PhD

Acknowledgements & Collaborators

Rice University

• Naomi Halas, PhD

• Jennifer West, PhD

• Lee Hirsch, PhD

• Scott Sershen, PhD

University of Texas - Austin

• Tinsley Oden, PhD

• Kenneth Diller, PhD

University of Texas – San Antonio

• Yusheng Feng, PhD

Biotex, Inc

• Ashok Gowda, PhD

• Roger McNichols, PhD

• Anil Shetty, MD, ME

• Kevin Pham

Nanospectra Biosciences, Inc

• Don Payne

• Jon Schwartz, PhD

Interventional Radiology MRI

• Kamran Ahrar, MD

• Judy Ahrar, MD

• Sanaz Javdi, MD

• Yvette Valenzuela, RT(R)MR

Neurosurgery BrainSUITE™• Jeff Weinberg, MD

• Van Luu, RT(R)MR

Urology

• John Ward, MD

Veterinary Medicine & Surgery

• Peggy Tinkey, DVM

• Rajesh Uthamanthil, DVM, PhD

• Agatha Bourne, DVM, PhD

• Sherry Klumpp, PhD

Work supported in part by: NIH CA79282, NIH AG19276, NIH CA101573, NIH 5T32CA119930, NSF CNS-0540033, NSF IIS-0325550, NSF IIO-0548741

The diseases which medicines cannot

cure, excision cures: those which excision

cannot cure, are cured by the cautery; but

those which the cautery cannot cure, may

be deemed incurable.

- Hippocrates Aphorisms (400 BCE)

Thank you for your time!

Email: jstafford@mdanderson.org