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
l
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 ([email protected])
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