dosimetry in digital mammography - linköping university · 2014-04-07 · dosimetry in digital...

Dosimetry in digital

mammography

Professor David Dance

NCCPM, Royal Surrey County Hospital, Guildford, United kingdom

Outline• Why do dosimetry?

• History

• Essentials of European breast

dosimetry protocol

• Practical breast dosimetry

• Uncertainties and limitations

• Key points

Why do dosimetry?

• To control the dose

– no higher than necessary

– BUT high enough to ensure that adequate image quality is obtained

– essential part of quality control

• optimise equipment selection, design and use

• To estimate risk

– comparison of risk and benefit

History

UK DOSE SURVEY 1981

• 5 cm phantom

• Entrance surface dose – TLD

• Dose range 0.9 to 45 mGy (!)

• kV range 24- 49 kV

• HVL range 0.2 mm Al to 1.7 mm Al!

• Entrance dose is a poor measure

of risk!

Transmission through 5 cm breast

0.001

0.01

0.1

1

15 20 25 30 35

Adipose

Gland

Energy keV

Tra

nsmission

Alternative dose measures

• Mid-breast dose

• Mean breast dose

• Average glandular dose(Mean glandular dose)

– Karlsson in 1976

– Recommended by ICRP in 1987

Essentials of European

breast dosimetry protocol

Original dosimetry

requirements in UK

• AGD: risk related dose measure

• Easy to implement

• Applicable for patient and phantom dosimetry

• Standard phantom for QC

– Equivalent to a typical compressed breast

Average glandular dose• Cannot measure AGD on patients

• Incident air kerma K (without backscatter) can be easily measured or estimated

• Need conversion coefficients which relate K to AGD

• Used in: IPSM Report 59 1989

European Protocol 1996

AGD = K g

Focal spot

Compression plateSimple breast phantomBreast support etc

Calculation of g

using Monte Carlo model (Dance 1990)

Calculate energy deposited in breast tissues

Simple breast modelHammerstein 1979

• 5 mm adipose shield region

• Central region with mixture of glandular and adipose tissues

• Fraction by weight of glandular tissue in central region is known as the glandularity

Hammerstein et al suggested the use of 50% glandularity for dosimetry

g-factors• g-factor is for 50% glandularity

• In principle g-factors depend upon

target/filtration and kV

Simplification• Tables of g-factor just based on HVL & breast

thickness

• For spectra in use at that time worked within ±

5%

g-factors

0

0.2

0.4

0.6

0.8

2 4 6 8

HVL 0.3 mm

HVL 0.6 mm

Breast thickness cm

g-fa

ctor

mGy/m

Gy

Improved formulation in 2000

Dance et al 2000, 2009, 2010, 2014

• Real breasts don’t all have 50%

glandularity – factors calculated for

0.1-100%

• Wider range of spectra used

AGD = Kgcs

• c corrects for glandularity

=1 for 50% glandularity

• s corrects for X-ray spectrum

=1 for Mo/Mo

• Used in:

– UK (IPEM report 89 2005)

– European Guidelines (latest

supplement is 2013)

– IAEA Code of Practice (TRS457)

AGD = Kgcs

c-factorsTabulated against HVL, breast thickness

and glandularity

0.30 0.35 0.40 0.45 0.50 0.55 0.60

HVL (mm Al)

0.6

0.8

1.0

1.2

1.4

c-fa

ctor

0.1%

25%

50%

75%

100%

5 cm breast

But what glandularity should be used?

Breast glandularity (UK)

s-factors

S Max ErrorMo/Mo 1.000 3.1%

Mo/Rh 1.017 2.2%

Rh/Rh 1.061 3.6%

Rh/Al 1.044 2.4%

W/Rh 1.042 2.1%

W/Ag 1.042 4.6%

W/Al s-table needed

“Standard breasts”

• For quality control and inter-system comparison need a simple phantom or phantoms

• Cheap and easily reproducible (PMMA)

• Should be approximately equivalent to typical compressed breast(s)

Breast/PMMA equivalence

• Same incident air kerma at PMMA

surface and breast surface

– Obtained by Monte Carlo simulation

– Validated by measurements with

breast tissue equivalent phantoms

PMMA for standard breast

dosimetry

PMMA mm 20 30 40 45 50 60 70

Breast mm 21 32 45 53 60 75 90

Gland’y % 97 67 41 29 20 9 4

PMMA equivalence age 50-64Dance et al, 2000

Practical breast dosimetry

European Guidelines

Fourth Edition

Supplements 2013

Method:

Section 2b2.3

g,c and s factors:

Appendix 5

Practical dosimetry using

blocks of PMMA (6 monthly)

Objective is to simulate exposure of ‘standard breasts’

• Stage 1: determine exposure settings using blocks of PMMA

• Stage 2: determine incident air kerma & HVL

• Stage 3: calculate AGD

• Stage 4: review results

Stage 1 - Exposure parameters

for blocks of PMMA

• Use usual clinical

exposure settings

• Record mAs, kV

and target/filter

• Repeat for 20, 30,

40, 45, 50, 60, 70

mm blocks

Exposure parameters for blocks

of PMMA – Notes I

• PMMA blocks

– Thickness correct within ± 0.5 mm

• PMMA and breast equiv thicknesses are

different

– If AEC works on thickness, add expanded

polystyrene spacer to adjust total thickness (do not

cover chamber with spacer)

– If AEC works on attenuation, spacer should not be

necessary

Exposure parameters for

blocks of PMMA – Notes II

• Apply standard compression

• AEC on mid-line and closest to chest wall

• If digital AEC looks at densest region of breast,

exposures may not be a good match to those

obtained clinically

Stage 2 - Determination of

incident air kerma and HVL

g, c and s depend upon the HVL

K and HVL are determined with the

compression paddle in place

AGD = Kgcs

Determination of incident

air kerma - I

• Incident air kerma at upper surface of PMMA– Measurement of tube output at a known

distance from tube focus

– Apply inverse square law to calculate air kerma at top surface of PMMA

• Measurement is made with paddle in contact with the dosimeter

Determination of incident

air kerma - II

• Use a calibrated dosimeter suitable for

mammography

• Position dosimeter on line joining focal spot

to a point on the mid-line of the breast

support 6 cm from chest wall edge.

Determination of incident

air kerma - II

• For dosimeters with backscatter rejection

position dosimeter on breast support

• For dosimeters without backscatter

rejection

position the dosimeter higher up to avoid or

reduce the effect of backscatter

Tables A5.1 & A5.2

give g and c vs

PMMA cm & HVL

Table A5.4

gives s

AGD = KgcsStage 3

Stage 4: review results

0

1

2

3

4

5

6

7

2 3 4 4.5 5 6 7

acceptable

achievable

PMMA thickness cm

AGD m

Gy

0

5

10

15

20

25

30

35

0.5 1 1.5 2 2.5 3

MGD (mGy)

Nu

mb

er

of

syste

ms

98% < 2.5 mGy

AGD: 53 mm breast (45 mm PMMA)

Patient dosimetry

• Measure tube output and HVL with paddle in place

• Record exposure parameters for patient series (e.g.50 or more)

– kV & target/filter

– mAs

– compressed breast thickness

• Calculate AGD

Patient dosimetry – Note

• Accuracy of displayed thickness

should be checked by applying a

typical force (e.g. 100 N) to rigid

material of known thickness

• Accuracy of ±2 mm or better

required

Tables A5.5 –A5.7

give g and c vs

breast cm & HVL

c for age 50-64 or 40-49

Table A5.4

gives s

AGD = Kgcs

Average patient AGD for different

digital systems: 2010 UK data –Oduko et al IWDM 2010

DR CR FS0.0

0.5

1.0

1.5

2.0

2.5

Image receptor type

AG

D (

mG

y)

Uncertainties & Limitations

Uncertainties in AGD using

PMMA (typical)

• Dosimeter calibration 5%

• AEC and dosimeter

reproducibility 2%

• Uncertainty in g due to

3% uncertainty in HVL 2%

• Chamber positioning (2mm) 1%

• Phantom thickness (0.5mm) 2%

• s-factor approximation 4%

• Overall 7%

Errors in using tabulated

equivalent PMMA thickness

• Exact PMMA thickness required for

equivalence varies with beam quality.

• Error is largest for highest kVs and

thickest breasts.

– 0-60 mm breast: up to 1.0 mm PMMA

– 80 mm breast: up to 1.7 mm PMMA

– 100 mm breast: up to 2.4 mm PMMA

Errors in AGD associated with non-

equivalence of tabulated PMMA

thickness (W/Al spectrum)

20 30 40 50 60 70 80 90 100

0

5

10

1530 kV 35 kV 40 kV

Breast thickness (mm)

Err

or

%

Air kerma measurement

• Measurements should be made with the dosimeter in contact with paddle (to mirror MC model)

• Scatter contributes e.g. ~7% to air kerma for 2.4 mm polycarbonate paddle

• Some semi-conductor dosimeters are collimated and will not record scatter

Correction is then needed (Hemdal, Rad Prot Dosim. 2011)

Uncertainties in patient AGD typical

and assuming breast model is correct

• Dosimeter calibration 5%

• Dosimeter etc. reproducibility 2%

• Uncertainty in g due to

3% uncertainty in HVL 2%

• Patient thickness (2 mm) –

impact on g and ISL 3%

• s-factor approximation 4%

• Overall, for one exposure 7%

PLUS Standard error on mean from sample of 50 patients about 7%

Some limitations of the model

• Heel effect ignored

• Simple breast geometry

• Breast composition data

(Hammerstein) based on

– 4 samples glandular tissue

– 8 samples adipose tissue

– 6 samples skin tissue

– results show considerable variability

in composition

Effect of adipose shield

thickness on g

2 4 6 80.95

1.00

1.05

1.10

1.15

1.20

2.5 mm adipose shield

Thickness (cm)

g /

g(s

tan

dard

)

Effect of glandular tissue

composition on g

2 4 6 80.85

0.90

0.95

1.00

1.05

1.10High O

Low O

Thickness (cm)

g /

g(s

tan

dard

)

“Real” distributions of

glandular tissue

• The distribution of glandular tissue in

patients will not be uniform and large

differences in conversion factors can be

expected.

• Modelled using breast phantoms from

Predrag Bakic (Dance et al, RPD 2005)

• Differences in AGD of up to ~40% found

European vs USA breast

model (Wu)

• Adipose shield replaced by 4mm skin– Different initial attenuation

– Different volume containing glandular tissue to share energy absorbed

• Unclear if dose from photons scattered in paddle included

• Different cross section compilation

• Differences of up to 16% in conversion coefficients

Key points

Key points

• The conversion factors are influenced by the model parameters.

• Whatever breast model is chosen for QC dosimetry, it need only be an approximate representation of the true situation. The present European and American models give standard procedures and are fine for QC.

• Large errors can occur if either model is used to estimate AGD for individual patients. But the models are fine for QC of patient doses.

Acknowledgements

• Recent work has been funded under the

EU 6th Framework Programme – the

HighRex project and the UK funded

OPTIMAM and OPTIMAM2 projects

• I am grateful to my colleagues at

NCCPM for provision of data, and for

discussions