Skin Model and its impact on Digital Mammography

1

Rodrigo T. Massera; Alessandra Tomal

Institute of Physics "Gleb Wataghin”

University of Campinas

Campinas, Brazil

Outline

2

Motivation

• Mammography

• Dosimetry – Mean Glandular Dose

Methodology

• Implemented Models

• How?

Results

• MGD vs Skin Model

• Differences

Conclusions• Summary

Introduction

3

Population-based screening programs

Use Ionizing Radiation

Quality Control and Optimization

Why is dosimetry important in

mammography?

Mean Glandular Dose (MGD)

Adipose

Tissue

Glandular

Tissue

Skin

Real Breast

Incident Photons

Adipose

Tissue

Glandular

Tissue

Skin

Real Breast

Energy Deposited: Glandular Tissue Directly measured?

Monte Carlo simulation

Mean Glandular Dose (MGD)

6

Parameters to consider...

• X-ray spectrum;

• Geometric Model;

• Breast Thickness;

• Breast Composition (𝑓𝑔);

• Skin Model?• 5 mm Adipose Tissue ( Dance 1990)

• 4 mm Skin Tissue (Wu 1991/Boone 1999)

Using breast-CT: ≈1.44 mm (Vedantham et al 2012)

≈1.45 mm (Huang et al 2008);

+ adipose layer

Previous Estimations:

Current Measures:

64% thinnerCredits: Boone & Hernandez 2016, AAPM. “Changing

Perceptions and Updated Methods for Mammography

Dosimetry”

Mean Glandular Dose (MGD)

Objectives

7

Study the impact of skin models on Mean

Glandular Dose in Digital Mammography

• Geometry

• MGD calculus

Adapt MC Code

• MGD X Skin ModelsAnalysis

Outline

8

Motivation

• Mammography

• Dosimetry – Mean Glandular Dose

Methodology

• Implemented Models

• How?

Results

• MGD vs Skin Model

• Differences

Conclusions• Summary

9

Monte Carlo code:

• PENELOPE (2014) + penEasy (2015)

Beam Parameters:

• Monoenergetic (8 – 60 keV)

• Polyenergetic (22 – 35 kV):

• Mo (Mo-Rh)

• Rh (Rh)

• W (Rh-Al-Ag)

*X-ray spectra from Hernandez et al (2014)

Methodology

Methodology

10*Compositions from Hammerstein et al 1979

Breast Model*

• t = 2 cm – 8 cm

• Glandularity (𝑓𝑔) = 1%-100%

11

Breast Model*

• t = 2 cm – 8 cm

• Glandularity (𝑓𝑔) = 1%-100%

Skin shielding Models

I. 5 mm adipose;

II. 4 mm skin;

III. 1,45 mm skin;

IV. 1,45 mm skin + 2 mm adipose;

V. 1,45 mm skin + 3,55 mm adipose;

*Compositions from Hammerstein et al 1979

Methodology

12

Adipose

Tissue

Homogeneous Glandular-

Adipose Tissue Mixture

Skin

Monte Carlo Simulations

How do we separate the deposited

energy between tissues?

penEasy: 𝐸𝑎𝑣𝑔

Mean Glandular Dose (MGD)

13

MGD Weighing method (Dance 1990)

Simulation starts:

𝐸𝑔𝑙𝑎𝑛𝑑=0

InteractionType

IncoherentPhotoelectric

(I)

(II)

(III)

Simulation Ends:

Return𝐸𝑔𝑙𝑎𝑛𝑑

MGD = 𝐸𝑔𝑙𝑎𝑛𝑑

𝑀𝑎𝑠𝑠 × 𝑓𝑔

nMGD = 𝑀𝐺𝐷

𝐾𝑎𝑖𝑟

Mean Glandular Dose (MGD)

14

Dosimetry

Air Kerma from Primary

PhotonsMGD

Code modifications...

nMGD = 𝑀𝐺𝐷

𝐾𝑎𝑖𝑟

~40.000 simulations

15

User Input

• Mono/Poly;

• Energy Range/Anode Filter;

• Breast Thickness range;

• Breast Composition;

• Skin Model;

• Detector Type;

• Antiscatter Grid Model;

• etc

Python

Script

List of Simulations

Parallel Simulations

Windows/Linux full compatibility

PENELOPE

Data Collection

and savingPython

Return Data

Automatization with Python™

~40.000 simulations

16

Automatization with Python™

3-10 min/simulation – Uncertainty (1 - 0.25%)

Processor i7 7700 3.6 Ghz

Outline

17

Motivation

• Mammography

• Dosimetry – Mean Glandular Dose

Methodology

• Implemented Models

• How?

Results

• MGD vs Skin Model

• Differences

Conclusions• Summary

Results - Code Validation

18

AAPM – Report 195 (2015)

19

• 5 cm thick;

• 20% 𝑓𝑔;

• 1.45 mm skin;

Results: Skin shielding models

20

Monoenergetic Beam

20

1% 𝑓𝑔 100% 𝑓𝑔

21

Monoenergetic Beam – Depth Dose 18 keV20% 𝑓𝑔

2 cm breast 8 cm breast

Results: Skin shielding models

22

Polyenergetic Beam

2 cm 8 cm

20% 𝑓𝑔

36%

15%

34%

16%

Results: Skin shielding models

23

Polyenergetic Beam Mo/Mo 28 kV

2 cm breast

21%

23%

Results: Skin shielding models

24

Polyenergetic Beam Mo/Mo 28 kV

21%

17%

Results: Skin shielding models

1% 𝑓𝑔

Results: Summary

25

Polyenergetic Beam – Skin Models

Outline

26

Motivation

• Mammography

• Dosimetry – Mean Glandular Dose

Methodology

• Implemented Models

• How?

Results

• MGD vs Skin Model

• Differences

Conclusions• Summary

Conclusions

27

𝑓𝑔 (≈50%) BreastThickness

(≈350%)

SkinModel

(≈40%)

MGD

Variation

Tube Potential(≈130%)

Depth Dose : skin attenuation and Homogeneous Mixture Volume

Anode/Filter(≈90%)

• Skin model affects the MGD up to 37%;

• Larger variations: low energies; high glandularity, thin breasts

• The Skin Model has a significant impact on MGD estimates;

• Reduce the uncertanties;

• Patient-specific dosimetry;

• Heterogeneous breast

Acknowledgement28

• Process 2016/15366-9

• Process 2015/21873-8 • Process 483170/2015-3

Rodrigo T. Massera Bruno L. Rodrigues

José Maria

Fernandez-Varea

Lab Members and AlumniCollaborators

Our Institution

29

University of Campinas (UNICAMP): 1st in Latin America

Credits: Lucas Rodolfo de Castro Moura -

http://www.lrdronecampinas.com.br/

Funded in 1966

30

Thank You!

atomal@ifi.unicamp.br

rmassera@ifi.unicamp.br

Rodrigo T. Massera

MSc Student

IFGW - UNICAMP