Latest Advances in Digital Breast Tomosynthesis

Wei Zhao, Bo Zhao and Yue-houng HuDept. of Radiology

State University of New York at Stony Brook

Limitation of Mammography

• False diagnoses due to superposition of breast tissue

• Solution: 3D imaging – Breast CT: Cone-beam CT (180°) with

dedicated prone table

– Breast tomosynthesis: modified mammo gantry with limited angular range (15o-60o)

Breast Tomosynthesis

• Fast clinical transition

• Limited angular range: 15~60o

• Slice thickness: 1mm

• Projections: 11- 50

• Dose: ~1 - 2 times single mammogram

Mammogram Tomosynthesis

Biopsy proven cancer Courtesy: B. Ren, Hologic

Different DBT DesignCompany System param. Views Detector Scan time

(s)Recon.

GE ±20°, step/shoot(MGH, Sunnybrook)

15 CsI/a-Si 100 um

15-23 MLEMSART

±30°, step/shoot(U. of Michigan)

21 7 s SART

Hologic ±7.5, continuousMultiple sites

11 a-Se, 70 um2x2 binning

10 FBP

Siemens ±22°, step/shootDuke,Malmo,SUNY

25 a-Se, 85 um 12.5/20 Bin/full

FBP

Dexela ±12-20°(U. of Virginia)

13 Fiber optic coupled CCD

30 s MLEM

X-counter ±13° 48 Gas counting, 48 slit, 60 um

FBPiterative

Sectra ±5.5º, 6-site trial 21 Si counting, 21 slit, 50 um

3-8 s

Factors Affecting Image Quality

• Detector performance

– DQE at low dose: 1/NView

– Temporal performance: lag and ghosting

• Imaging geometry

– Angular range, number of views

– Focal spot blur: continuous tube travel

• Reconstruction algorithm– Analytical: FBP (filter design)

– Iterative

Detector Performance

0 1 2 3 4 5 60.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

data-line direction

DQ

E

spatial frequency (cycles/mm)

nbin bin 0.43 mR 2.15 mR 6.01 mR

Imaging Geometry: spatial vs. frequency domain

Reconstructed slices

x

y

dxdy

dz

fz

fxz

Spatial domainFrequency domain

fzN Y

fxNY

rec tangular area ( )used for reconstruction

fz ,fxNY NY

space covered by acquired images voxel

θ

B. Zhao and W. Zhao, Med. Phys. 2008

3D Image quality metrics: MTF, NPS and DQE

Slice thickness filter: eliminate aliasing

fz

fxΘ

fz

fx

Aliasing fNY

2fNY

2fNY

Slice thickness 1 mm:Noise aliasing in z

Noise Measurement

40slices

Reconstructed 3d image

Projection noise image

x

zy

dxdy

dz

voxel

Focal spot

detector

4cm Lucite

Voxel dimension

dx=0.085 mm

dy=0.085 mm

dz=1 mm

3D FFT

B. Zhao et. al, Med. Phys. 2009

In-depth NPS

Ramp x Hann

fz

fx

experiment model

•Effect of angular range θ•Effect of reconstruction filters•Effect of 3D sampling, ST filter

2fNY

θ(

SBP

Ramp x Hann x ST (width =0.47cycles/mm)

fx

fy

fz

NPS vs. angular dose distribution

fx

fz

measurement

model

ADS-1

ADS-1

ADS-2

ADS-3

ADS-2

ADS-3

In-plane NPS

Measured 2D NPS from top slice

Measured 2D avg NPS from all slices

calculated 2D avg NPS

•Reconstruction: Ramp + Hanning

fx

fy

fx

fy Projection 2D NPS

3D NPS

-6 -4 -2 0 2 4 60

10000

20000

30000

40000

50000

60000

avg NPS NPS of top slice NPS from model

NP

S(A

DU

2 mm

2 )

freq(cycles/mm)

Oversampled Point Spread Function (PSF)

y

x (mm)

017(mm)

0

14

x

y

70 um wire tilted 20 degrees

z

z

PSF/artifact vs. Angular Range

+20o +15o +10o +5o

MTF – Dependence on Filters

0

0.5 cyc/mm 0 10

x (cyc/mm)

z

SBP

RA

RA+SA

RA+SA+ST

Impact of Angular Range on In-Plane MTF

• Increasing angular range improves MTF at low frequencies

0 2 40.0

0.5

1.0

1.5

2.0

2.5

3.0

MT

F

freq(cycles/mm)

angular range 10o

20o

40o

60o

180o

In-Plane MTF Measurement

20 40 60 80 100 120 140 160 180 2000

500

1000

1500

2000

2500

3000

3500

4000

ES

F

position(pixel)

Derivative(LSF)

|1D FFT|: MTFx

Reconstructed in-plane image

Focal spot

detector

2cm luciteEdge phantom

Edge phantom: 0.2mm Al

In-Plane MTF

0 2 4 6 8 100.0

0.2

0.4

0.6

0.8

1.0

MT

F

freq(cycles/mm)

Measurement Model

recon setting: Rampx Hann xST, prebin

Low f drop caused by ramp filter and limited angular range

High f drop caused by other reconstruction filters

fx

fy

Projection 2D MTF

In-plane MTF

Dependence on Angular Range

±20o ±10o

ACR phantom imaged at 28kVp, W/Rh, 1.7mGy

Recon: FBP with ST filter

SDNR=2.5 SDNR=1.4

Reconstruction filter designs

• Ramp + limited angle:loss of low frequency (breast density)

• Modified filters:recovers density, but increased out-of-plane blur

• Iterative recon: improvement at the cost of computation 0 1 2 3 4 5 6

0.0

0.5

1.0

1.5

2.0

ampl

itud

e

Spatial frequency (cycles/mm)

Slice thickness (fz)

ramp

Polynomial, similar to SART

apodization

J. Zhou et. al. Med. Phys. 2007; Ludwig et. al. IWDM2008; B. Ren, et. al. SPIE 2009

Reconstruction filter comparison

FBP with Ramp SIRT

Adapted from Ludwig et. al. (IWDM 2008)

Polynomial

Clinical Application of DBT

• Screening/diagnosis• Contrast enhancement: angiogenesis• Multimodality imaging:

– DBT + automated ultrasound (AUS)• P. Carson et. al. U. Michigan

– DBT + optical tomography• Boas et. al. MGH

– DBT + limited angle SPECT• M. Williams et. al., U. Virginia

Kinetics

0

1

2

3

0 2 4 6 8Time (minutes)

mg/c

m2S

igna

lp19_scout_enhan3

Post

Dynamic Contrast-Enhanced (CE) Tomosynthesis

- =

p19_scout_enhan3

Pre

Pre Post

X-ray projections

Tomo reconstruction

-

Subtraction (CE) Courtesy Dr. R. Jong

CE-Tomosynthesis

Mammo Tomo CE-Tomo

Courtesy M. Hill and J. Mainprize

• DS DBT system, GE Medical Systems

• Automated tube motion

• No anti-scatter grid

• HE:

– Rh target - 0.25 mm Rh + 0.25 mm

Cu filter

– 49 kVp

• LE

– Rh target – 0.25 mm Rh filter

– 32 kV

• 7 projections: 40° arc, 6.7° apart

• Dose per HE/LE data set ~single mammographic view (~2 mGy)

• Filtered backprojection

Temporal vs. Dual-Energy CE-DBT

Courtesy Ann-Katherine Carton

Case 1: poorly differentiated invasive ductal carcinoma

Carton A.-K., Chen. et al., Br. J. Rad. 2010Courtesy Ann-Katherine Carton

GE Gen II Research Tomo Unit60o angle21 projections7.5 sec

Dual-modality compression paddle

GE Logiq 9US system

Stationary digital x-ray detector

U. Michigan

Courtesy: Drs. P. Carson and M. Goodsitt

Automated US Scanning System

• Translator & transducer are flipped up out of field for tomo acquisition

• Translator & transducer flipped down for AUS acquisition

TPX paddle

xy translator

UStransducer

Courtesy: Drs. P. Carson and M. Goodsitt

Co-location Reader Study Display GUI showing Corresponding VOI’s in DBT & AUS

• AUS-suggested ductal extension

mammo tomo

InvasiveCancer

AUSCourtesy: Drs. Carson & Goodsitt

Further Development for DBT Systems

• Gantry:– Faster scan, more views

• Generator/tube:– dual-energy, shorter pulse/less focal spot

blur

• Detector:– Better low dose performance, faster readout

– photon-counting

New detectors: Improve low dose performance

-

S

MULTIPLEXER

SCANNING C

ONTROLHigh gainphotoconductor

HV biaselectrode

pixelelectrode

TFT

storagecapacitor

G

D

x-rays

-

S

MULTIPLEXER

SCANNING C

ONTROL

Avalanchea-Se (HARP)

pixelelectrode

TFT

storagecapacitor

G

D

StructuredCsI

x-rays

Direct FPI:Photoconductor with higher gain

Indirect FPI:Avalanche photodetector

Summary

• Physics of DBT: 2D�3D– Optimization of acquisition and recon.

• New clinical applications:– Contrast enhanced CEDBT– Multimodality imaging

• Challenges for further system development

Acknowledgements

• Financial support: – Siemens AG

– Army Breast Cancer Research Program

– NIH

• Contributing slides:– Drs. Baorui Ren, James Mainprize, Mitch

Goodsitt, Ann-Katherine Carton, Mats Danielsson