Get Complete Project Material File(s) Now! »

## Breast equivalent material for VBD computation

As described in Section 1.2.2, most of the V BD computation methods for mammographic projection images are based on calibrations using a breast equivalent material available from a single manufacturer following the method of White [150] and Fatouros [51] (CIRS Inc., Norfolk, VA). Consequently this breast equivalent material is the cornerstone for estimating breast densities from mammographic X-ray images. A fundamental requirement for the phantom material is a close match in attenuation properties between phantom material and real breast tissue (precision) in the energy domain of mammography (i.e. 15 keV to 30 keV). The material should in addition allow consistent measurements within the samples (homogeneity) and from sample to sample of the same nominal glandular content (consistency).

From another point of view, all breast density assessment methods should be validated on patients against available imaging modalities capable to identify the different tissues and with the possibility to measure the volume of the breast and fibroglandular tissue. CT is a 3-dimensional imaging modality with a good geometric accuracy. It can effectively distinguish fibroglandular tissue from adipose tissue, based on attenuation characteristics of both tissues and with the possibility to quantify volumes, which makes it a good candidate for the validation of breast density quantification methods [50]. We discuss this method in Chapter 4. In the frame of our work, it was therefore of particular interest to double check the properties of the calibration phantoms using CT as well.

### CT measurements of the breast equivalent phantom

The capability of CT to measure the attenuation properties of the breast equivalent phantoms must first be verified and the accuracy must be determined. The breast equivalent material generally used in mammography is available under the form of rectangular plates.

The influence of the shape and configuration of these plates on the CT measurements has to be studied. Once the use of dual energy CT is verified, we can use the voxel values, expressed in Hounsfield Units (HU), in the images of the plates. They can be used to assess the homogeneity of attenuation properties and then the consistency between different plates of the same nominal breast density equivalence. The linearity of HU to the nominal breast density equivalence can be evaluated. Finally linear attenuation coefficients can be derived from the mono-energetic HU provided by dual-energy CT as described by Wu [155] and used by Van Gompel [63] and Zhang [159]. The results for these coefficients are compared with values from the literature [66, 84].

#### Breast equivalent phantom characteristics

The differences for the eight middle volumes in Configurations 1 and 2 varied between 0.1 and 1.3 HU. This must be compared to an average standard deviation of 4 HU within the ROIs constituting the volumes. The standard deviation over the mean values of these ROIs is 1 HU. The difference of the average HU of the volumes of the two configurations a different orientation is therefore not significant. The orientation of the plates can be neglected in the center of the plates and in the middle of the stack. The two outer plates of the stack of small plates differed between the Configurations 1 and 2 by 4.1 and 6.0 HU. Therefore these plates were not considered for further tests for the verification of the phantoms. The problems are caused by the sharp edges of the phantoms, so no problems are expected for patients.

The standard deviations over the ROIs respectively in water and in the plates in Configuration 4 were comparable: 0.6 in water, 0.8 and 0.6 in the two adipose equivalent plates and 0.7 and 0.7 in the two fibroglandular equivalent plates. Since the variations from ROI to ROI were not significantly different from the variations within water, the homogeneity of the plates is confirmed. Some more important moderate local deviations were observed in some image planes (maximum difference to average value of 1.5 HU within water, 2.4 HU and 5.5 HU in the two adipose equivalent plates and 2.0 HU and 2.1 HU in fibroglandular equivalent plates respectively), but do not contradict the homogeneity assessment. The standard deviations over the ROIs of the same nominal breast density equivalence in Configuration 3, neglecting the two outer plates, were 1.7 HU for the adipose and 1.5 HU for the fibroglandular equivalent plates. Maximum differences versus the average values of the plates were 2.9 HU for the adipose and 3.2 HU for the fibroglandular equivalent plates. The average HU of the nominal breast density equivalence of 0% and 100% were respectively -67 HU (standard deviation 1.7%) and 55 HU (standard deviation 4.5%).

The measurements of the HU in the small plates in Configuration 1 were plotted against their nominal breast density equivalence (0%, 30%, 50%, 70% and 100%) in Figure 2.2. The HU were measured in the center of the plate in 157 consecutive slices in total. The deviating values for 0% and 100% nominal breast density equivalence came from the outer plates, where the consistency of the CT measurement is not guaranteed (see above). A least square fit showed the good linear correlation between the nominal breast density equivalence and the HU as expected from Equation 2.1 (R2 = 0.989).

**Breast equivalent phantom in mammography**

The breast equivalent phantom can be used for the calibration measurements of our model of the volumetric breast density. We used the large 24×18×2cm3 and 24×18×1cm3 plates with attenuation equivalent to 0% and 100% fibroglandular tissue. In projection mammography the density in a point is the integrated density along the path of the X-ray arriving in that point. This is an advantage for our density computation method, since a phantom with a density different from 0% or 100% can be created by combining them. To confirm this we investigated the influence of the order of the plates. We did this for a phantom with a total thickness of 4 cm and 25% density. This was reached with a 1 cm-plate of 100% density and a 1 and 2 cm-plate of 0% density. Different acquisitions were performed at the same exposure values while inverting the order of the plates.

**Table of contents :**

List of abbreviations

Abstract

R´esum´e

Pr´esentation du projet

Densit´e volumique du sein

Etat de l’art

Nos contributions

Energie d´epos´ee dans la glande

Etat de l’art

Nos contributions

Conclusion

Samenvatting

Probleemstelling

Volumetrische borstdensiteit

State of the art

Onze bijdragen

Energie geabsorbeerd door het klierweefsel

State of the art

Onze bijdragen

Conclusie

**1 Introduction **

1.1 General problem statement

1.2 State of the art

1.2.1 Breast anatomy and histology

1.2.2 Breast density

1.2.3 Breast dose

1.2.4 Conclusion

1.3 Research objectives

I BREAST DENSITY

**2 Validation of breast equivalent phantom material **

2.1 Breast equivalent material for VBD computation

2.2 CT measurements of the breast equivalent phantom

2.2.1 CIRS

2.2.2 Configurations

2.2.3 Measuring the CIRS characteristics

2.3 Breast equivalent phantom characteristics

2.4 Breast equivalent phantom in mammography

2.5 Conclusion

**3 Breast density computation **

3.1 Volumetric breast density: introduction

3.2 VBD: from theory to practice

3.2.1 Theoretical model for the computation of V BDMX from digital mammographic images

3.2.2 Implementation of the V BD computation

3.2.3 Application to mammographic images

3.3 Calibration of the VBD computation

3.3.1 VBD for phantoms

3.3.2 VBDMX for the database of mammographic images

3.4 Conclusion: VBD computation is possible

**4 New validation of the breast density computation**

4.1 CT versus mammography

4.2 VBD computation in CT

4.2.1 Theoretical derivation

4.2.2 Database of mammographic and CT images

4.3 VBDCT versus VBDMX

4.3.1 Calibration of the CT method

4.3.2 VBDCT for the databases

4.3.3 Correlation between V BDCT and V BDMX

4.4 Same breast, same VBD

Discussion Part I

II BREAST DOSIMETRY

**5 Validation of the Monte-Carlo simulation tool CatDose **

5.1 Monte-Carlo simulations

5.2 TG 195 AAPM manual

5.3 Computation of dose conversion factors

5.3.1 Dance conversion factor

5.3.2 Boone conversion factors

5.4 Limits and strengths

**6 Evaluation of irradiation in mammography **

6.1 Dosimetry for individuals

6.2 Use the right density

6.3 New quantity used for individualized risk

6.4 Computation of the local GIE

6.5 Discussion and conclusion on individual risk assessment

**7 Segmentation of glandular tissue from tomosynthesis **

7.1 Brief introduction to tomosynthesis

7.2 Tomosynthesis limitations

7.3 Segmentation method

7.4 Reconstructions

7.5 Textured phantom

7.6 Real patient cases

7.7 Conclusion on tomosynthesis segmentation

Discussion Part II

**8 General conclusions and perspectives **

APPENDIX

A Breast statistics

B Level estimators

List of publications

**Bibliography**