Attenuation Correction in Hybrid MR-BrainPET Imaging

Transcription

1 Mitglied der Helmholtz-Gemeinschaft Attenuation Correction in Hybrid MR-BrainPET Imaging Elena Rota Kops Institute of Neuroscience and Biophysics Medicine Brain Imaging Physics

2 Interactions of 511 kev Photons in Matter Photoelectric effect Compton scattering I = I 0 exp( - µ dx) µ µ compton + µ photoelectric I 0 = initial photon flux I = photon flux without interaction in medium x = thickness of the medium µ = linear attenuation coefficient (it depends on the atomic number Z of the material and on the photon energy) µ compton µ photoelectric µ Half-values thks (1/cm) (1/cm) (1/cm) (cm) Brain tissue ~0.096 ~ ~ Bone ~0.169 ~0.001 ~ Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 2

3 Linear attenuation coefficient µ in PET scanner Emission The detector pair measures: The measured counts: Counts m = A dx exp( - µ dx) Counts corr = Counts m exp( + µ dx) = A dx Line of Response (LOR) = Counts m ACF = A dx ACF = attenuation correction factor Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 3

4 In case of the good old dedicated PET scanners Blank Transmission I o I o rotating line source with positron emitter Ge-68 I during the blank scan I = I o during the transmission scan I = I o * exp( - µ dx I 0 = exp(+ µ dx) = ACF I Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 4

5 Attenuation image I 0 ln = µ dx) I which is exactly the Radon transform to get an image of the attenuation coefficients µ Line integrals of the transmission image µ dx => forward projection I 0 = ACF for each LOR I Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 5

6 Attenuation Measurement in PET/CT CT (HU) smoothed CT (HU) µ-map (1/cm) 68 Ge measured µ-map (1/cm) CT 0HU CT > 0HU µ µ PET PET = µ = µ PET H 2 O PETO H 2 PET 1 PET 1 with µ H 2 O = cm and µ Bone = cm CT 1 CT 1 µ H 2 O = cm and µ Bone = cm ( CT ) / 1000 CT 2 ( PET PET µ 2 + CT H O µ Bone µ H O ) CT 1000( µ µ ) Bone CT H 2 O cm cm 1 1 Burger C et al. PET attenuation coefficients from CT images: experimental evaluation of the transformation of CT into PET 511-keV attenuation coefficients. Eur J Nucl Med (2002) 29: Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 6

7 Attenuation Measurement in PET/CT CT (HU) smoothed CT (HU) µ-map (1/cm) 68 Ge measured µ-map (1/cm) CT BreakPoint µ PET = 9. 6x10 5 ( HU ) cm 1 CT > BreakPoint µ PET = a ( HU ) + b cm 1 with a and b functions of kvp Carney JPJ et al. Method for transforming CT images for attenuation correction in PET/CT imaging. Med Phys (2006) 33: Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 7

8 From PET/CT to MR-PET Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 8

9 Attempts to get a segmented MRI Two-Compartment Method ACF = exp(+ µ (x,y)dl') LOR brain thks skull thks ACF (brain+skull) ACF (only brain) rel. difference (cm) (cm) (1/cm) (1/cm) (%) LOR 1 LOR 1 LOR 2 LOR LOR 4 LOR 4 LOR 3 LOR LOR 5 LOR 5 Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 9

10 Attempts to get a segmented MRI Morphological Operators MRI CT segmented MRI segmented CT Dogdas B, Shattuck DW, and Leahy RM Segmentation of Skull and Scalp in 3-D human MRI Using Mathematical Morphology Human Brain Mapping (2005) 26: Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 10

11 Attempts to get a segmented MRI Neural network- and Knowledge-based segmentation First results T1-weighted MPRAGE data set Classification and Segmentation based on - Neural network and - Knowledge on relative positions into GM, WM, CSF, extracerebral region Wagenknecht et al., Methods Inf Med 43(2004), Further segmentation ( ongoing work) Brain tissue, different cavities, skull, scalp Wagenknecht G Central Institute for Electronics (ZEL), Jülich In collaboration with Rota Kops E, Herzog H, INB3, Jülich Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 11

12 Attempts to get a segmented MRI Two-points Dixon MRI. Whole Body Water Fat Martinez-Möller A, Souvatzoglou M, Navab N, Schwaiger M, and Nekolla S MR-based attenuation correction for whole-body MR/PET J Nucl Med (2008) 49 (Supplemet):65P Segmented image CT Lesion SUV = - Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 12

13 Attempts to get a segmented MRI ultra short TE (UTE) Normal tibia Difference UTE image: IR=500/0.08 minus 4.5/250 ms Frontal bone Difference UTE image: IR=500/0.08 minus 4.5/200 msec at 1.5T Tyler DJ, Robson MD, Henkelman RM, Young IR, and Bydder GM Magnetic Resonance Imaging with Ultrashort TE (UTE) Pulse Sequences: Technical Considerations J Magn Reson Imaging (2007) 25: Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 13

14 Attempts to get a segmented MRI ultra short TE (UTE): New Results Keereman V, Vandenberghe S, De Deene Y, Luipaert R, and Broux T MR-based Attenuation Correction for PET Using an Ultrashort Echo Time (UTE) Sequence Presented at the IEEE Medical Imaging Conference 2008, Dresden Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 14

15 Alternative methods to get attenuation maps Template Based. Theory + = T1-MR template (e.g. a self-build one) individual T1-MR Transmission template (e.g. a self-build one) individualized attenuation map complete individualized attenuation map reformatting into ECAT7 image and voxel dimensions forward projection reconstruction with with attenuation correction and ROI analysis complete individualized attenuation map image: 128x128x65 voxel: 5.15 x 5.15 x 2.42 mm 3 sinogram of ACFs Rota Kops E, and Herzog H Alternative Methods for Attenuation Correction for PET Images in MR-PET Scanners IEEE NSS/MIC Conference record (2007) pp & Template-basierte Schwächungskorrektur für PET-Bilder in MR-PET-Geräten Nuklearmedizin (2008) 47:A63 Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 15

16 Alternative methods to get attenuation maps Template Based. Results Measured attenuation image Female and Male Data 8.00 Male subject relative Differenzen (%) Frontal cortex Parietal cortex Occipital cortex Caudate nucleus Putamen Thalamus Female subject Template based on SPM2 with template based fp attenuation factors y = 0.93x R 2 = 0.95 Female subjects y = 1.00x R 2 = with measured fp attenuation factors w ith measured attenuation factors with tamplate based fp attenuation factors y = 0.89x R 2 = 0.91 Male subjects y = 0.99x R 2 = with measured fp attenuation factors with measured attenuation factors Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 16

17 Alternative methods to get attenuation maps Template based. Ongoing Work. Brain Ongoing work: 1) Optimization of templates: possible gender specific differences and 2) Use of UTE sequences on 3T. Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 17

18 Alternative methods to get attenuation maps MR-CT Database. Atlas Database of corresponding MR-CT Pairs {MR i } {CT i } {MR new } Registration for every MR i to the patient MR new and apply to CT i Find all neighbouring MR patches Perform Gaussian Regression on patch and position CT estimates for every voxel Scale CT values (HU) to PET attenuation values Estimate PET Attenuation Image Hofmann M et al, MR-based PET Attenuation Correction Method and Validation Presented at IEEE NSS/MIC Conference 2007 & MRI-Based Attenuation Correction for PET/MRI: A Novel Approach Combining Pattern Recognition and Atlas Registration J Nucl Med (2008) 49: Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 18

19 Alternative methods to get attenuation maps MR-CT Database. Results x 10 7 Relative difference between PET (MR-AC) and PET (CT-AC) Activity in PET (MR-AC) Nr of segmented VOI Activity in PET (MR-AC) x 10 7 Fig. 3: Left: Relative differences between activity concentrations in the PET image obtained using MR-based AC in comparison to CT-based AC. Right: Scatter plot, with the ideal diagonal in green. Hofmann M et al, MR-based PET Attenuation Correction Method and Validation Presented at IEEE NSS/MIC Conference 2007 & MRI-Based Attenuation Correction for PET/MRI: A Novel Approach Combining Pattern Recognition and Atlas Registration J Nucl Med (2008) 49: Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 19

20 Alternative methods to get attenuation maps Ongoing Work. Brain Extensive evaluation on 17 MR-CT pairs, 3 PET/CT-MR triplets with standard T1SE sequence on 1.5T. Ongoing work: 1) Optimization of Flash 3D and 2) Use of UTE sequences on 3T. MR Image Predicted PseudoCT Real CT PseudoCT vs CT prediction error of on average 84.7 HU per voxel. On a dataset of 17 MR-CT pairs, the average PseudoCT prediction error was HU. Whole Body Ongoing work: 1) Development on a probabilistic framework. 2) Optimization of dual echo protocol with less than 30s acquisition time per bed position, i.e. breath hold acquisition possible. Hofmann M et al, MR-based PET Attenuation Correction Initial Results for Whole Body Presented at IEEE NSS/MIC Conference 2008 in Dresden Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 20

21 Summary Two possible ways to get ACF: - Segmentation of the clinical MR images -- Differentiation between skull and cavities --- Knowledge based segmentation --- Usage of different MR pulse sequences (UTE) - Alternative methods -- based on measured transmission images (only brain) -- based on CT images (brain and whole body) Oct. 28, 2008 INB Medicine (Brain Imaging Physics) Folie 21