1. Motion Correction on Dual ECG and Respiratory Gated 3D Cardiac PET/CT Data
Mohammad Dawood1,3,4, T Kösters2, M Fieseler1,3, F Büther1,4, X Jiang3, KP Schäfers1,4
1European Institute of Molecular Imaging, University of Münster
2Institute for Computational and Applied Mathematics, University of Münster
3Institute of Computer Science, University of Münster
4University Hospital Münster

2. Data: Hybrid PET/CT
PET
PET/CT CT

3. Problem: Motion during data acquisition
PET, Minutes
CT, Seconds

4. Problem: 1. Attenuation correction
CT: Static
PET: Respiratory
phases
Expiration
Inspiration

5. Problem: 2. Image blur
Extent of motion
Superimposed image

6. Gating + Motion Correction
Clinical relevance
Motion artifacts lead to errors:
wrong „staging“ of the tumors (Osman et al 2003, Erdi et al 2004)
„Uptake Values“ are incorrect (Nakamoto et al 2004)
small tumors may remain undetected (Papathanassiou et al 2005)
Solution?
Gating + Motion Correction

7. Dawood et al, Medical Physics, Sep. 2007
Solution: Part 1 - Gating
Dawood et al, Medical Physics, Sep. 2007

8. Method: Amplitude vs. Time-based gating
Respiratory Signal
Gate 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6
Time
Gates
… possible with Listmode Acquisition
Amplitude-based Gating
Time-based Gating
Dawood et al, Medical Physics, Sep. 2007

9. Method: Amplitude- vs. Time-based gating
P1 Amplitude based
P1 Time based
P2 Amplitude based
P2 Time based

10. Brightness consistency constraint I(x,y,z,t) = I(x+∂x,y+∂y,z+∂z,t+∂t)
Solution: Part 2 - Optical flow
Brightness consistency constraint
I(x,y,z,t) = I(x+∂x,y+∂y,z+∂z,t+∂t)
I.V = -It
I = Grey value
V = Flow
x,y,z,t = Position

11. IxnVx+IynVy+IznVz = -Itn
Solution: Part 2 - Optical flow
Lucas/Kanade: Optical flow is locally consistent
Ix1Vx+Iy1Vy+Iz1Vz = -It1
IxnVx+IynVy+IznVz = -Itn
Local consistency

12. Horn/Schunck: Optical flow is globally smooth
Solution: Part 2 - Optical flow
Horn/Schunck: Optical flow is globally smooth
∫((I.V+It)+(|u|2+|v|2+|w|2))dxdydz
Brightness consistency
Global smoothness

13. Dawood et al, IEEE Trans Med Imaging, August 2008
Solution: Part 2 - Optical flow
Lucas/Kanade + Horn/Schunck + Discontinuity Preservation
∫ (ψ1(I.V+It)+αψ2(w,|u|2+|v|2+|w|2)) dxdydz
Global smoothness
Local consistency
Discontinuity preservation
Ψ2 (w,s)
Dawood et al, IEEE Trans Med Imaging, August 2008

14. Remaining Motion ~ 0.3 mm << 1 Voxel
Results: Visual
PET-Data before correction
-8 gates
Remaining Motion ~ 0.3 mm << 1 Voxel
PET-Data after correction
-8 gates

15. Remaining Motion ~ 0.3 mm << 1 Voxel
Results: Quantitativ
Remaining Motion ~ 0.3 mm << 1 Voxel

16. + Dual ecg and respiratory gating and motion correction
Respiratory motion, 6 gates
Combined motion, 36 gates +
Cardiac contraction, 6 gates

17. Results: Visual Combined motion, 36 gates
Combined motion, 36 gates, motion corrected
After motion correction, 36 gates

18. Lamare et al, Phys Med Bio, 2007
Data-based motion correction and reconstruction
Data-based approach outperfoms the image-based motion correction
Line of Response
Lamare et al, Phys Med Bio, 2007

19. Attenuation correction: PET-derived dynamic CT-based mu-map
Dynamic CT-based mu-maps derived from a static mu-map
and motion information from the PET

20. Dual ecg and respiratory gating and motion correction
Target Gate Ungated Motion Corrected

21. Results: Myocardial thickness
One gate 12,5 mm
Before motion correction 17,2 mm
After motion correction 12,5 mm

22. Results: Myocardial thickness

23. Thank you!