Iterative techniques for metal artifact reduction · 2011. 9. 28. · reducing metal artifacts in...

Post on 12-Oct-2020

2 views 0 download

transcript

h f lIterative techniques for metal artifact reductionartifact reduction

F. Edward Boas andDominik Fleischmann,Stanford Radiology

MDCT conference, San Francisco2011‐06‐132011 06 13

CT metal artifactsCT metal artifacts

CT metal artifactsCT metal artifacts

X‐ray source Low error

High error, due to:• Poisson noise• Beam hardening• Scatter• Motion

Metal deletion technique (MDT)Metal deletion technique (MDT)Use all of the data to reconstruct the metal pixels

… but only use non‐metal data to reconstruct non metal pixels

X‐ray source Detector array

the metal pixels … reconstruct non‐metal pixels.

Metal

Soft tissueSoft tissue

Metal deletion technique (MDT)Metal deletion technique (MDT)

Use forward projection iteratively to replaceUse forward projection iteratively to replace detector measurements that involve metal.

Simulated scansBoneWaterSoft tissue

FBP LI MDT

Iron

Exact

Poisson noise

Beam hardening

Motion

All errors

Simulated scansBoneWaterSoft tissue

FBP LI MDT

Iron

Exact

Poisson noise

Beam hardening

Motion

All errors

Simulated scansBoneWaterSoft tissue

FBP LI MDT

Iron

Exact

Poisson noise

Beam hardening

Motion

All errors

Simulated scansBoneWaterSoft tissue

FBP LI MDT

Iron

Exact

Poisson noise

Beam hardening

Motion

All errors

Cholecystectomy clips: FBPCholecystectomy clips: FBP

Cholecystectomy clips: LICholecystectomy clips: LI

Cholecystectomy clips: MDTCholecystectomy clips: MDT

Embolization coils: FBPEmbolization coils: FBP

Embolization coils: LIEmbolization coils: LI

Embolization coils: MDTEmbolization coils: MDT

Dental fillings: FBPDental fillings: FBP

Dental fillings: LIDental fillings: LI

Dental fillings: MDTDental fillings: MDT

Hip replacements: FBPHip replacements: FBP

Hip replacements: LIHip replacements: LI

Hip replacements: MDTHip replacements: MDT

Clinical scans  FBP LI  MDT 

Clips 

   

 Coils 

   

 Dental fillings 

   

 Hip replacements 

Image quality rank (raw data)Image quality rank (raw data)

1MDT LI FBP SART

Best 1

rank

1.0 ± 0.0 *Best

11 scans178 slices

2

e qu

ality 

2.1 ± 0.1 *

178 slices

Average ± S.E.M.

3rage im

ag

2.1 ± 0.1 * statistically significant difference

Aver

3 5 ± 0 1 3 5 ± 0 14

3.5 ± 0.1 3.5 ± 0.1Worst

Metal artifact reduction from DICOMMetal artifact reduction from DICOM

If raw data is not available it can be simulatedIf raw data is not available, it can be simulated by forward projecting DICOM files generated by 

the scannerthe scanner.

Improved image quality (DICOM)Improved image quality (DICOM)

%

80%

ed Small implants

50%

60%

70%

es im

prove Small implants

Large implants

30%

40%

50%

e of im

age

10%

20%

ercentage

0%MDT Rubout Rubout plus Wei

Pe

80 slices.  Data from Caroline Golden, Sam Mazin, et al.

Improved diagnosisImproved diagnosis

FBP MDT

Rectal cancerHip replacement

Improved diagnosisImproved diagnosis

FBP MDT

Apparent tip of the pacer wire

Improved diagnosis (DICOM)Improved diagnosis (DICOM)

FBP MDT

 

Stroke     Aneurysm coil 

Decreased resolution near metalDecreased resolution near metalHorizontal resolution Vertical resolution

Metal

Decreased resolution near metalDecreased resolution near metalResolution near a 10 mm metal implant

1.4

1.6

m)

0.8

1

1.2

ution (m

m

0 2

0.4

0.6

yresolu

0

0.2

0 10 20 30 40 50 60

Distance from metal (mm)

ConclusionsConclusions

1 MDT reduces metal artifacts due to Poisson1. MDT reduces metal artifacts due to Poisson noise, beam hardening, and motion.

2 MDT has better image quality than other2. MDT has better image quality than other techniques (p=0.0005), and may change the diagnosisdiagnosis.

3. MDT works on a variety of scans, from hip l i ireplacements to moving pacer wires.

AcknowledgementsAcknowledgements

Dominik Fleischmann (Stanford)( )Roland Bammer (Stanford)

Caroline Golden (National University of Ireland Galway)Sam Mazin (Stanford)Jon Abelson (Stanford)Chris Beaulieu (Stanford)

Christoph Panknin (Siemens)

Funding: None

Thank you!Ed Boas (boas@stanford.edu)

www.revisionrads.comwww.revisionrads.com

Key references:

Boas FE and Fleischmann D. (2011) “Evaluation of two iterative techniques for reducing metal artifacts in computed tomography.”  Radiology, doi: 10.1148/radiol.11101782.

Golden C, Mazin SR, Boas FE, Tye G, Ghanouni P, Gold G, Sofilos M, Pelc NJ. (2011) “A comparison of four algorithms for metal artifact reduction in CT imaging.”  SPIE Medical Imaging Conference 2011, Orlando, Florida.g g g g f , ,

Boas FE. (2011) “Iterative reduction of artifacts in computed tomography images using forward projection and an edge‐preserving blur filter.”  U.S. Patent ApplicationPatent Application.

Extra slidesExtra slides

Metal deletion technique (MDT)Metal deletion technique (MDT)

Adaptive detector element sizeAdaptive detector element size

7

8

35

40

ze

Exact photon countActual detected photons (with Poisson noise)

5

6

7

25

30

35

lemen

t si

oton

s

Actual detected photons (with Poisson noise)Adjusted photon countAdaptive detector element size

3

4

5

15

20

25

detector e

ber of ph o

1

2

3

5

10

15

Ada

ptive d

Num

0

1

0

5

0 10 20 30 40 50

A

Detector element number