26
Effects of Slice Thickness Filter in Breast Tomosynthesis Filtered Back Projection Reconstruction Linlin Cong 1 , Weihua Zhou 2 , *Ying Chen 1,2 1 Biomedical Engineering Graduate Program 2 Department of Electrical and Computer Engineering, Southern Illinois University, Carbondale, IL 62901 (* Corresponding Author) IEEE GENSIPS 2011 Medical Imaging Laboratory 1

Slice profile ieee2011_siu

  • Upload
    linlinc

  • View
    871

  • Download
    0

Embed Size (px)

DESCRIPTION

 

Citation preview

Page 1: Slice profile ieee2011_siu

1

Effects of Slice Thickness Filter in Breast Tomosynthesis

Filtered Back Projection Reconstruction

Linlin Cong 1, Weihua Zhou 2, *Ying Chen 1,2

 1Biomedical Engineering Graduate Program

2Department of Electrical and Computer Engineering, Southern Illinois University, Carbondale, IL 62901

(* Corresponding Author)

IEEE GENSIPS 2011

Medical Imaging Laboratory

Page 2: Slice profile ieee2011_siu

2

OUTLINE

Introductions

Breast Tomosynthesis Imaging

System

Tomosynthesis Image

Reconstruction and Simulation

Results

Conclusions

Page 3: Slice profile ieee2011_siu

3

Introduction

Page 4: Slice profile ieee2011_siu

4

Breast Cancer Breast Cancer:

• Most common cancer among women worldwide;

• Second leading cause of cancer related death among women.

Symptoms:• No symptom in the early stage, regular

breast exams are important;• Breast lump or lump in the armpit,

change in the size, shape, fluid from the nipple.

• Once the patient is diagnosed with breast cancer, next step is staging (Grade: 0, I, II, III, IV). Higher the grade is, poorer the outcome of the treatment will be.

Page 5: Slice profile ieee2011_siu

5

Mammography

Traditional mammography:• Currently, a standard and important clinical

screening and diagnosis for early detection of breast cancer;

• Cheap, low radiation dosage. Limitations of traditional mammography:

• 20% false negative rate, many call backs from screening;

• low positive predictive value, about 30% of breast cancers are still missed in mammography;

• 2D imaging system, difficult to distinguish a cancer from overlapped breast tissues.

Page 6: Slice profile ieee2011_siu

6

Breast Tomosynthesis

3D slice images provide depth information

Improve conspicuity of structure by removing the visual clutter associated with overlying anatomy

Promising to reduce recall rates, and to increase cancer detection accuracy.

Low dosage; relatively cheap Extensive attentions from academic

communities and industrial vendors have been paid to this promising field.

Page 7: Slice profile ieee2011_siu

7

Breast Tomosynthesis Imaging System

Page 8: Slice profile ieee2011_siu

8

Siemens Mammomat Inspiration

GEIMS GIOTTO Hologic Selenia Dimensions

Examples of current commercial breast tomosynthesis prototype systems :

Page 9: Slice profile ieee2011_siu

9

Most of these systems re-utilize the traditional partial isocentric mammography design.

X-ray tube moves in an arc across the breast

Series of low dosage images are acquired at different angles

Limitation is X-ray tube’s movement may introduce motion blur and cause patients’ discomfort.

Page 10: Slice profile ieee2011_siu

10

A Novel Nanotechnology Enabled Digital Breast Tomosynthesis

Prototype System Invented by our collaborators at the

University of North Carolina Chapel Hill Built up with fixed multi-beam field-

emission X-ray sources, no movement of X-ray tubes;

Total scanning time: about 11.2 seconds for typical 25 projection views

Advantages:• No motion blur• Less scanning time, so decreasing the

waiting time

Page 11: Slice profile ieee2011_siu

11

A novel multi-beam x-ray source developed by Zhou Lu et al.

• Fixed multi-beam field-emission x-ray (MBFEX) sources based on unique properties of carbon nanotube electron emitters.

• The total scan time for a typical 25 views is about 11.2 seconds.

Page 12: Slice profile ieee2011_siu

12

Tomosynthesis Image Reconstruction and

Simulation

Page 13: Slice profile ieee2011_siu

13

BREAST TOMOSYNTHESIS IMAGING SYSTEM SIMULATION

The image acquisition system we used to get the projection images was simulated based on the parallel imaging system.

• 25 X-ray sources.

• The path of tubes is parallel to the plane of detector.

• Two sets of data were simulated to investigate the filter effects:

Page 14: Slice profile ieee2011_siu

14

The image acquisition system:

Page 15: Slice profile ieee2011_siu

15

IMAGING SYSTEM: Data 1 One Sphere:

• placed at 30mm above the detector • radius = 5mm.

Page 16: Slice profile ieee2011_siu

16

IMAGING SYSTEM: Data 2 Two overlapping spheres were simulated.

• Sphere 1: height = 20mm above detector, radius = 5mm• Sphere 2: height = 40mm above detector, radius = 10mm

Page 17: Slice profile ieee2011_siu

17

Reconstruction Algorithm

• Mathematic Reconstruction Methods: Shift and Add (SAA) Backprojection(BP)

• Filter-based Reconstruction Methods: Filtered Backprojection(FBP) Matrix Inversion Tomosynthesis (MITS)

• Statistical Reconstruction Methods: Maximum Likelihood Expectation

Maximization(MLEM)

• Algebraic Reconstruction Methods: Simultaneous Algebraic Reconstruction

Technique (SART)

Page 18: Slice profile ieee2011_siu

18

Reconstruction Algorithm

• Mathematic Reconstruction Methods: Shift and Add (SAA) Backprojection(BP)

• Filter-based Reconstruction Methods: Filtered Backprojection(FBP) Matrix Inversion Tomosynthesis (MITS)

• Statistical Reconstruction Methods: Maximum Likelihood Expectation

Maximization(MLEM)

• Algebraic Reconstruction Methods: Simultaneous Algebraic Reconstruction

Technique (SART)

Page 19: Slice profile ieee2011_siu

19

Filtered Back Projection (FBP) Ramp Filter:

• To suppress the component in low x frequency(x)

• strengthen high frequency (x) part.

Low-pass Filter(Hamming Window):• To suppress noise

Page 20: Slice profile ieee2011_siu

20

FBP: Profile Filter

To reduce the effects caused by discontinuation of the z-border with a sharp step function

Appled along the Z direction Equation:

Hprofile_filter

Page 21: Slice profile ieee2011_siu

21

Results

Page 22: Slice profile ieee2011_siu

22

Impulse Responses Analysis

a) Impulse response without profile filter

b) Impulse response with profile filter

c) Intensity profiles of figure (a) d) Intensity profiles of figure (b)

Page 23: Slice profile ieee2011_siu

23

Single Sphere Data

a) Reconstructed image of 30mm plane without profile filter

c) Reconstructed image of 30mm plane with profile filter

b) Vertical plane of the object along Z-direction without profile filter

d) Vertical plane of the object along Z-direction with profile filter

e) Intensity profiles of figure (a) f) Intensity profiles of figure (b)

Page 24: Slice profile ieee2011_siu

24

Two Overlapping Sphere Data

a) Reconstructed image of 20mm plane without profile filter

b) Reconstructed image of 40mm plane without profile filter

c) Reconstructed image of 20mm plane with profile filter

d) Reconstructed image of 40mm plane with profile filter

e) Vertical plane of objects along Z-direction without profile filter

f) Vertical plane of objects along Z-direction with profile filter

Page 25: Slice profile ieee2011_siu

25

CONCLUSIONS

Effects of Profile Filter:• Enhance the sharpness

• Reduce the ringing artifacts

• Make the reconstructed objects spread out more uniformly along the depth (z) direction

• Reduce the mutual interference between objects located on the neighboring slices

Page 26: Slice profile ieee2011_siu

26

ACKNOWLEDGMENT

• We thank our collaborators at The University of North Carolina at Chapel Hill (UNC) and group members at Southern Illinois University (SIU).

• The related work has been supported by Southern Illinois University and U.S. National Institutes of Health (NIH/NCI R01 CA134598-01A1).

Medical Imaging Laboratory