Dose calculations for

Brachytherapy using TG-43

Alex Rijnders
Europe Hospitals
Brussels, Belgium
a.rijnders@europehospitals.be

AFRA Training course, Rabat, Morocco, July 2009

Introduction
Despite large dose gradients:

Inaccurate dose calculation

for an excellent implant
procedure
May be as bad as

Accurate dose calculation

for a terrible implant
procedure

Introduction (continued)
We need to improve our dose
calculation technique as we are
developing
the
implant
procedures.

A. Meigooni

Recommended BT
accuracy:
Physics global accuracy 5-10%
Input data and algorithm TPS numerical
accuracy of at least 2%
AAPM TG56, 1997

Current most commonly

algorithm:
AAPM Task Group 43: Brachytherapy
Dose Calculation Formalism
Nath et al., Med.Phys. 1995
Update1: Rivard et al., Med.Phys. 2004
U1Supplement1: Rivard et al, Med.Phys. 2007

Previously: Sievert Integral

Sievert 1921
Cassell 1982
Williamson 1988

Simplified model
L

P(r, ) or
P(x,y)

Y
I(x, y)

eq

Ly

Ra t

'
t sec
d
2 e
1

r
t

'

'
dl

Simplified to point source

water
water

distance d
point source

water

=
D water

.
K ref

en

1
d2

air

. (d)

Quantities:
water

en

air

Ratio of mean mass energy absorption coefficients water to air

(almost independent of energy, and therefore almost always equal to
ratio of mean mass energy transfer coefficients, except for low
energy)

water

=
D water

K ref

en

1
d2

. (d)

air

Quantities:
1
d2

Inverse square law term, relative to the distance at

which the reference air kerma is defined (e.g., at 1
cm, or at 100 cm)

water

=
D water

K ref

en

1
d2

. (d)

air

Quantities:

(d)
Correction factor for scatter and attenuation at
distance d from the source
Generally based om Meisberger (1968) data
Often applied as polynomial function

depth dose, dose as a function of distance for linear Ir-192

sources with the same linear source strength

Ir-192w
ireso
u
rced
o
serate

doserate (cGy/uur)

1-8cm 1m
u
G
y
/h@1m,p
ercm

1.2

1cm

1.0

2cm
8cm

0.8

3cm

0.6

4cm

5cm

0.4

6cm

0.2
1cm

0.0
0

7cm

distance (cm
)

10

8cm

Sievert limitations
Does not take into account real scatter behaviour

s, f: mathematical best fit, not physical

quantities
=> Acceptable results for 137Cs and 192Ir, but
errors up to 25% for 125I (AAPM 1995)

How to Calculate Seed Implant Dosimetry?

CT

radiograph

TG-43 Basic Approach

Calculate (Monte-Carlo) and measure the
dose distribution around a source
Parameterize TG43 parameters to fit to the
measurements
(TG43-Algorithm)-1
Experimental

TG43 parameters
TG43-Algorithm

Patient

TG-43 Basic Approach

Determine the dose distribution around a
source in standard conditions using at least 1
(preferable 2) reported experimental
measurements and 1 Monte Carlo study,
done by independent investigators.
Consensus data derived from these data
by an authority, eg AAMP TG43 or
ESTRO/Braphyqs

TG-43 Cylindrically symmetric

source

BT Dose Calculation: TG43

G(r, )
D ( r , ) Sk
g (r ) F (r, )
G (r 0, 0)
.

TG-43
Brachytherapy Dosimetry Formalism

G(r, )
D ( r , ) Sk
g (r ) F (r , )
G (r 0, 0)

D(r , )

SK

g(r)
G(r,)
F(r,)

dose rate to water at point P(r,

source strength, numerically = Reference Air
Kerma Rate
dose rate constant
radial dose function
geometry function
2-D anisotropy function

Brachytherapy Dose Calculation

G(r, )
D ( r , ) Sk
g (r ) F (r , )
G ( r 0, 0)

Sievert:

Dose rate constant

D ( r 0, 0)

Sk

Ratio of the dose at the reference

position over the source strength

= > Converts Air Kerma

to dose at the reference
point

Radial Dose function

Dose fall-off along the transverse axis of the source
(absorption and scatter effects in water)

G(r, )
D ( r , ) Sk
g (r ) F (r , )
G (r 0, 0)

g (r )

D ( r , 0) G ( r 0, 0)

D ( r 0, 0 ) G ( r , 0)

Geometry function
Deals with inverse square law, eliminating
largest variation in other parameters

1
G(r , ) 2
r

Point source approximation

Line source approximation

G(r , )
Lr sin
1
G(r , ) 2
r L2 / 4

Line source, = 0

Calculation Algorithm
Y
P( x, y) or P( r,)

Brachytherapy
Source

Anisotropy function
Accounts for anisotropy of the dose distribution around the
source, including absorption and scatter in source and water

G(r, )
D ( r , ) Sk
g (r ) F (r, )
G (r 0, 0)

F (r , )

D(r , ) G (r , 0)

D(r , 0) G (r , )

Anisotropy function

F (r, )

Anisotropy function:
2D approximation

an(r )

Anisotropy factor:
1D approximation
(source orientation unknown)

an

Anisotropy constant:
Use no longer recommended!

Anisotropy
function

Anisotropy factor

Length L

Summary I:
It is a Two-Step 2D - Calculation Method

(r0, 0): (1.0cm, 90)

2004: Revised AAPM TG-43

BT Dosimetry Formalism (2-D)

GL ( r , )
D ( r , ) Sk
gL ( r ) F ( r , )
GL ( r 0 , 0 )

D(r , )
SK

dose rate to water at point P(r,

air kerma strength

dose rate constant

gL(r)radial dose function (for line source appr.)
GL(r,)

geometry function (line source approximation)

F(r,)

2-D anisotropy function

Revised AAPM TG-43

BT Dosimetry Formalism (1-D) BEST:

GL ( r , )
D ( r , ) Sk
gL ( r ) an ( r )
GL ( r 0 , 0 )

D(r , )
SK

dose rate to water at point P(r,

air kerma strength

dose rate constant

gL(r)radial dose function (for line source appr.)
GL(r,)

geometry function (line source approximation)

an(r)

1-D anisotropy function

Comparison of 1D formalisms

GOOD

r0
D ( r , ) Sk 2 gp ( r ) an ( r )
r

BEST

GL ( r , )
D ( r , ) Sk
gL ( r ) an ( r )
GL ( r 0 , 0 )

Comparison of 1D formalisms
r 02
D ( r , ) Sk 2 gL ( r ) an ( r )
r

GL ( r , )
D ( r , ) Sk
gp ( r ) an ( r )
GL ( r 0 , 0 )

BAD

BAD

GOOD

r0
D ( r , ) Sk 2 gp ( r ) an ( r )
r

G (r , )
D ( r , ) Sk
g ( r ) an ( r )
G (r )

BEST

0,

Data entry in the TPS

TG43 data can be entered either :
As lookup tables (using linear interpolation)
As mathematical model fit: polynomial, other
Often data supplied by the TPS manufacturer
In some systems accessible to the user, in
others closed

Data entry in the TPS

Possible problems linked to this:
Extrapolation beyond published data (to 0, larger
distances)
Polynomial fitting might not always give the best
result, attention for behaviour outside the range used
for fitting

Mathematical fitting
The Original TG43 recommended
g(r) = ao + a1r + a2r2+ a3r3+ a4r4+ a5r5
Double exponential fit suggested by Furhang and
Anderson:

g(r) = C1 er C2 er

Modified polynomial suggested by Meigooni et al :

5th order Polynomial fit of g(r) vs Double

exponential and Modified polynomial fit

1.2

- - - - - - 5th order polynomial fit

Modified Polynomial

Radial Dose Function, g(r )

Radial Dose Function, g(r )

1.2

0.8
0.6
0.4
0.2

- - - - - - 5th order polynomial fit

Double-Exponential fit

1
0.8
0.6
0.4
0.2
0

0
0

Distance (cm)

10

Distance (cm)

10

Where to find data?

Preferentially use consensus data sets

Low photon energy sources:

AAPM TG43U and Supplement
RPC registry (rpc.mdanderson.org)
ESTRO website

www.estro.org/estroactivities/Pages/GEC-brachytherapycommitteeactivity.aspx

High energy sources:

ESTRO website (+ www.uv.es/braphyqs)
AAPM: work in progress

Literature

Where to find data?

Attention not to mix up data:
As geometry factor interferes in calculated g(r)
and F(r,), same geometry function should be
used in TPS
Check final result against the published dose
rate tables

Be aware !
Although Sievert approach looks like TG43
Factors from one formula cannot be used
in another formula !!
E.g. do not combine TG43 F(r,) with Sievert
integral dose calculation!

Other approaches (TPS)

Dose rate tables stored for each source
Cylindrical symmetry might be
assumed
Interpolation errors if geometry
function (inverse square law) is
not removed

Source specification

KR

Reference Air Kerma Rate: to be used in

-Calibration certificate
-Dose rate table
-TPS
-Prescription
-Reporting

Unfortunately
Use of Activity is still encountered
-Danger: use of different Air Kerma Rate
constant (or exposure rate constant) at the time
of measurement (manufacturer) and in the TPS

Cs: from 3 to 3.31

137

192

Ir: from 4.0 to 5.0

Source specification in TPS

Some allow any specification
Others allow exclusively Activity

=> Aa

or

Ac

ApparentContained
A and constants introduced in TPS must be
the same as those used to obtain A from
Reference Air Kerma Rate

Source decay

dose rate

T1/2 = 30
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00

Permanent implant

20

40
t

60

80

time (h)

Dose = dose rate x time

(mathematically: integration over time)

100

Source decay
For short time implantations (HDR, PDR),
or long lived isotopes: dose rate can be
considered constant
In case of afterloaders: decay handling
either by TPS or by Afterloader (or both)
For manual-LDR (Ir) implants: compensate
for decay during treatment (TPS-manually)
Permanent implants: integration over time

Point source Line source

For small sources, with no
anisotropy
=> Point source

Seeds, but orientation not known

=> Line (Point) source, 1D anisotropy

HDR PDR sources:

=> Line source, 2D anisotropy

Short distances from a linear

source
=> Line source, but even then dose on
the source encapsulation ????

Line source approximation

Number of point sources
Number of elementary
line sources
Line source model of
correct length
=> Only last method can correctly model the anisotropy at
close distance or along the source axis, but even then dose at
surface of source not correctly calculated

Line source approximation

Curved sources have to be
decomposed in linear
segments

Limitations of TG43 algorithm

Line source cylindrical source

Homogeneous water patient
Full scatter patient
Transit dose (for afterloaders)
Intersource effect
Applicators
Shielding

Cylindrical source
Geometry function should be source
(design) dependant
=> Change of TPS structure

Does not effect accuracy as corrected for in

g(r) and F(r,)

Lack of heterogeneity corrections

High energy sources: nearly the same
behaviour in tissues involved as in water
Low energy sources: importance of photoelectric effect increases as energy decreases

I-125 g(r), Variable and Composition

125

I radial dose function, g(r)

1
water, p=1.00
tissue, p=1.00
tissue, p=1.05
tissue, p=1.15
tissue, p=1.25
tissue, p=1.50

0.1

0.01

0.001

0.0001
0

10

15
20
radius [cm]

25

30

Lack of heterogeneity corrections

Historically: no density data available for
BT planning, distance factor considered as
predominant
With the increased use of CT data:
increasing interest to incorporate
heterogeneity corrections
=> new algorithms (MC)

Lack of heterogeneity corrections

For HDR-PDR: at first glance problem could be
biggest in bronchial implants
But: Prescription done at a fixed distance from the
applicator (1 cm), dose effect correlated with this
prescription system, dose gradient over distance
far more important (palliative treatment).
Prasad 1985: 125I implant in lung: difference of 9 to
20% with dose to water

Lack of full scatter

Most TPS assume infinite and full scatter
conditions
Not true for some interstitial implants, close to the
skin: breast implants, skin, lip,
Mangold et al.(2001): skin dose in breast implants
up to 14% overestimated by TPS (TLD)
Bernard et al. (2005): skin dose in breast implants
up to 20% overestimated by TPS (MC)
Also shielding creates lack of scatter: 2 to 15%
dose reduction when using shielded vaginal
applicator

Transit dose
Source entry,
interdwell movements,
exit
Effect depends on:
Interdwell velocity
Source Intensity
Implant geometry
Prescribed dose

Transit dose
Bastin 1993:
Endobronchial BT, 4 fractions: dose to nasal
cavity, posterior pharynx and trachea: 58 cGy
Rectal/prostate template (18 needles), 4 fractions:
dose to subcutaneous tissue in center of implant:
68 cGy
=> This is to tissues (far) outside the treatment
volume, assumed to receive a negligible dose

Intersource effect
Depending on number of sources, composition,
geometry
AAPM 1997: typical prostate implant with large
number of 125I seeds: peripheral dose reduction up
to 6%
Perez 2003: Tip of tandem of 137Cs Selectron:
reduction more than 20%

Applicators
Still often metallic applicators, surrounding the
source cylindrically
E.g. interstitial needles (breast implant), 192Ir:
about 1% absorption,
Fletcher type applicator, 137Cs: about 6%
Could be taken into account during calibration (if
always same kind of applicators is being used),
but needs thorough experimental verification

Shielding
Often used in vaginal applicators to protect rectum,
urethra and/or bladder
Reduction of bladder-rectum dose of 6% to 50%,
depending on material and dimension of shield and
isotope
Some TPS do not allow corrections, some
implemented 1D correction, others a 2D correction
table (for a 3D problem)
Warning for OR-dose reporting

Message
Be aware off/take into account limitations of your
system/corrections needed
Whenever changes (improvements) in calculation
algorithms are implemented

=> Discuss the influence of these changes

with the radiation oncologist

Implementation of algorithms
What about TG43U recommendations for
Data Entry in Planning Systems
Seven systems reviewed:
Varian : VarisSeed, BrachyVision
Prowess Planning systems: 2D, 3D
ADAC/Philips: Pinnacle p3
Nucletron : Theraplan, SPOT
Williamson 2004

Data Entry: General Observations:

Not all of the systems reviewed provide full support for
TG43U1 data specification and formulary
Some require manipulation of TG43U1 style data to fit the
calculation models to achieve TG43 formulation
Thus data entry becomes a significant QC/QA issue.
One may need to combine TG43U1 style data for a given
source into surrogate functions to enter into a given
treatment planning system.
In this case, clear documentation is recommended.

Conclusions:
Shortcomings of current algorithms
Tissue heterogeneity corrections generally not available,
nor lack of full scatter correction
Shielding effects not accurately taken into account

Linear Source calculations

TG43 formulation was originally intended for short
brachytherapy sources, few mm in length
Elongated source extensions to TG43 needed (AAPM task
group)

Shortcomings of current algorithms (2)

Point Source calculations
Point source based distribution calculations are common particularly
where source center location but not 3D orientation is known and
where orientations are assumed to be randomly distributed.
1D anisotropy corrections simply scale the transverse radial dose
distribution in isotropic (spherical) geometry.
Linear source models provide more accurate anisotropy in single
source dose distributions and for ensembles of implanted sources.
Fixed geometry implants, including ribbons and plaques, lend to linear
source (TG43 2D formula) models