PET Module

  • Upload
    tolla

  • View
    38

  • Download
    2

Embed Size (px)

DESCRIPTION

PET Module. Ana Beatriz Solana, MS Qu Tian (Teresa), MS Instructor: Dr. Charles Laymon. Cocaine Study . Literature: Cocaine-Dopamine - Decrease dopamine release - Toxic for animal DA neurons [ Volkow et al. 1997; Martinez et al. 2007, 2009 & 2011; Seiden et al. 1987] - PowerPoint PPT Presentation

Citation preview

Slide 1

PET ModuleAna Beatriz Solana, MSQu Tian (Teresa), MSInstructor: Dr. Charles Laymon

Cocaine Study

Literature: Cocaine-Dopamine - Decrease dopamine release - Toxic for animal DA neurons [Volkow et al. 1997; Martinez et al. 2007, 2009 & 2011; Seiden et al. 1987]

Biological plausibility: Chronic cocaine use loss of dopaminergic terminals Inconsistent results due to - The time since last use of cocaine - The reliable assessment VMAT2 - [11C]DTBZ

Cocaine users dopamine transporter Why DTBZ? Whats the hypothesis

The features of VMAT2(1) Active in pre-synaptic vesicular membranes(2) Is involved in the transport of various monoamines, such as serotonin, norepinephrine, dopamine (from the cytoplasm to their storage vesicles)(3) VMAT2 in the striatum is reported to largely (95%) represent storage vesicles in the dopaminergic terminals.

In vitro, In vivo, [11C] NPA Narendran et al. JPET 333: 533-539, 2010[11C] Raclopride Narendran et al. Synapse 65: 1344-1349, 2011 [11C] DTBZ Narendran et al. Am J Psychiatry 169: 55-63, 20122Cocaine Study: VMAT2 - [11C] DTBZ

Objective To compare VMAT2 in 12 cocaine abusers vs. 12 healthy controls Method In vivo, [11C] DTBZ nondisplaceable binding potential using kinetic analysisAcquisition protocol

Hypothesis: Voxel-wise analysis would be valid as ROI approach. Specifically, [11C] DTBZ BPND would be lower in cocaine abusers than matched healthy controls in subregions of striatum as shown in previous ROI analysis. [11C]DTBZArterial blood sampleMRI Scan 1.5TTransmission scan(10 min)Emission data collection (60 min)

PETTo compare VMAT2 availability in a group of 12 recently abstinent cocaine-dependent subjects and matched healthy comparison subjects. Transmission scan is to obtain attenuation factors.After that, the radiotracer was injected in bolus of 20 seconds. Framing, time activity curveArterial blood sample 3Inter-frame Motion Correction** Pxmod software **SPM8 softwareNormalization:**PET -> MR templateVoxel-wise group analysis: cocaine vs. control **Voxel-wise2-tissueCompartmentModel

Voxel-wiseSRTM2Model

Voxel-wiseSRTMModel

FULL MODEL*SIMPLIFIED MODELS*Analysis PipelineThis is less noisy, but constrained under the assumptionsone-tissue compartment fitting well with the reference (if not fitting well, it indicates the reference is not well chosen, it may have some specific binding potential) andthe delivery from plasma/vascular to the ROI and the reference is the same (R1)

Blood is only availble for 20 subjects.Voxel-wise could be very noisy, 2-T model needs to obtain 4 parameters, whereas simplified models only need to estimate 2 or 3 parameters, but constrained by the assumptions. They could be good b/c they are less noisy. 4Inter-frame Motion CorrectionNormalization:PET -> MR templateVoxel-wise group analysis: cocaine vs. control Voxel-wise2-tissueCompartmentModel

Voxel-wiseSRTM2Model

Voxel-wiseSRTMModel

FULL MODELSIMPLIFIED MODELSAnalysis PipelineThis is less noisy, but constrained under the assumptionsone-tissue compartment fitting well with the reference (if not fitting well, it indicates the reference is not well chosen, it may have some specific binding potential) andthe delivery from plasma/vascular to the ROI and the reference is the same (R1)

Blood is only availble for 20 subjects.Voxel-wise could be very noisy, 2-T model needs to obtain 4 parameters, whereas simplified models only need to estimate 2 or 3 parameters, but constrained by the assumptions. They could be good b/c they are less noisy. 5Inter-frame Motion correctionSubjects can move over the time of the scanMotion correction is needed to compute voxel Time Activity Curve

Time Activity CurveF1 (15s) F2(15s) F3(15s) F9 (2m) F10 (5m) F11(5m) F18(5m) F19(5m) F20(5m)

We want to make sure each voxel across the frames is the sameAs you can see, the images had different contrasts across the data acquisition. The frame 106Inter-frame Motion correction1. Take one or the sum of a few consecutive frames w/o motion as the reference.

2. Sum the first 1-4 or 1-5 frames as the composite initial frame.

Frame 1(15s) Frame 2(15s) Frame 3(15s) Frame 4 (15s) Frame 5 (60s)

Frame 10 Frame 9-10 Frame 9-10-11Woods, J Comp Assit Tomography 19986. 1-4 or 1-5 have very low signals. Sum 1-4 or 1-5 (initial reference) coregister this initial reference to step 5 obtain the transformation matrix apply transformation matrix to 1, 2, 3, 4, 5, individually

73. Apply correction to all frames to the reference. 4. Apply transformation matrix from the composite initial frame to initial frames individually.5. Check motion corrected images.

Before correction After correction

Inter-frame Motion correctionWoods, J Comp Assit Tomography 1998When we apply transformation, we basically move the images to do interpolation which causes the smoothing.So that you can see, the corrected images are more smoothed than the uncorrected one.

3. All frames apart from the initial frames.For the initial frames. We apply8Analysis PipelineInter-frame Motion CorrectionNormalization:PET -> MR templateVoxel-wise group analysis: cocaine vs. control Voxel-wise2-tissueCompartmentModel

Voxel-wiseSRTM2Model

Voxel-wiseSRTMModel

FULL MODELSIMPLIFIED MODELSThis is less noisy, but constrained under the assumptionsone-tissue compartment fitting well with the reference (if not fitting well, it indicates the reference is not well chosen, it may have some specific binding potential) andthe delivery from plasma/vascular to the ROI and the reference is the same (R1)

Blood is only availble for 20 subjects.Voxel-wise could be very noisy, 2-T model needs to obtain 4 parameters, whereas simplified models only need to estimate 2 or 3 parameters, but constrained by the assumptions. They could be good b/c they are less noisy. 9PET2-Tissue Compartment Model

Arterial input functionA clearance parameter from vasculature to brain

A fraction moving to the specific compartment C2A fraction of the radiotracer diffusing back to plasma If the specific binding is reversible, a fraction of the radiotracer transferring back to C1Binding potential (BP) (nondisplaceable) =

Tracer is present in the arterial plasma (input curve Cplasma) and in two tissue compartments C1 and C2. The model configuration is sequential: Tracer is taken up with rate constant K1 into compartment C1. A fraction of it diffuses back to plasma with rate constant k2, another fraction moves further to compartment C2 with rate constant k3. Unless tracer is irreversibly trapped in the C2 compartment (k4=0), transfer back to the intermediate compartment is also going on with rate constant k4. The usual interpretation is that C1 represents free and non-specific bound tracer in tissue (non-displaceable compartment), and C2 represents specifically bound or metabolized tracer.

The arterial blood is used as a best approximation to the radioactivity concentration in capillaries at blood-brain barriere. the k3 parameter is reflective of the rate of specific binding (it is a composite measure that include kon the first order radioligand association rate constant for binding (since at high specific activity)6e. k4 is a clearance parameter and this is typically reflective of the radioligand dissociation rate from binding

a combined measure of the density of "available" neuroreceptors and the affinity of a drug to that neuroreceptor.

10

Simplified Reference Tissue Model &Simplified Reference Tissue Model 2 Lammertsma and Hume, 1996; Wu and Carson, 2002Assumption 2:

Distribution volume: ROI = Ref Assumption 1:

ROI & Ref follow a 1T

NOTE:- Both bias- SRTM < SRTM2 (k2)

a1+ C 2If the tracer kinetics in the target region are such that it is difficult to distinguish between free and specific compartments, the reference tissue model can be simplified further. This corresponds to the situation where the timeradioactivity curve of the region of interest can be fitted satisfactorily to a single tissue compartment model with plasma input, without signifi- cant improvement when a two-tissue compartment model is used.Note: In cases where the 1-tissue compartment model assumption does not apply, BP estimates from SRTM have found to be biased [43].

Note: The reference methods MRTM2 and SRTM2 require k2' as an input parameter. The k2' resulting from the SRTM method above is a suitable estimate. Therefore, when switching in PKIN from the SRTM model to the SRTM2 or MRTM2, k2' is automatically copied from SRTM, as long as Model conversion in the Configuration menu is enabled.

SRTM2 is designed specifically for voxel-wise analysis. Wu and Carson [32] aimed at making the Simplified Reference Tissue Model (SRTM) more robust for pixel-wise applicationsThis corresponds to the situation where the timeradioactivity curve of the region of interest can be fitted satisfactorily to a single tissue compartment model with plasma input, without signifi- cant improvement when a two-tissue compartment model is used.11SRTM & SRTM2 Lammertsma and Hume, 1996; Wu and Carson, 2002Assumption 1: Both ROI and Reference tissue follows a 1T model RegionModelVTChiSquareAICR2Ref.1Tissue 5.5795.73236.8130.963Ref.2Tissue6.1830.332-13.8260.998ROI1Tissue 22.0450.217-26.7930.997ROI2Tissue22.0120.246-19.8170.997Assumption 2: The delivery from plasma to the receptor-rich region and the reference region is the same (K1/k2 = K1/k2)In many cases, however, one or both regions might have kinetics more con- sistent with a 2T model. The three parameters used by SRTM could allow it to do a better job fitting such inconsistent data, and possibly produce small biases. By fixing one parameter in SRTM2, however, there is no flexibility in this method to adapt to data that is inconsistent with the 1T model. Therefore, it was important to assess whether model parameters, particularly BP, are more biased with SRTM2 than SRTM.

C1 with non-specific bound ligand or free ligand C2 with specific bound ligand In the simplified models, it assumes the binding process in non-specific and specific regions happens very quickly and cannot distinguish them.The reason why ROI fits 1-T model better is exactly the assumption tells us. k2 apparent is going to have the contribution in the specific region.The reason why reference fits 2-T model better 12Single voxel TAC and model fitX(56),y(55),z(33)BPND= 3.49Taking one single voxel TACFitting the curveBPND= 3.49X(56),y(55),z(33)BPND= 3.68BPND= 3.68

SRTM2

SRTMThe choice of reference region in positron emission tomography (PET) human brain imaging of the vesicular monoamine transporter 2 (VMAT2), a marker of striatal dopamine innervation, has been arbitrary, with cerebellar, whole cerebral, frontal, or occipital cortices used. To establish whether levels of VMAT2 are in fact low in these cortical areas, we measured VMAT2 protein distribution by quantitative immunoblotting in autopsied normal human brain (n=6). Four or five species of VMAT2 immunoreactivity (75, 55, 52, 45, 35kDa) were detected, which were all markedly reduced in intensity in nigrostriatal regions of patients with parkinsonian conditions versus matched controls (n=9 to 10 each). Using the intact VMAT2 immunoreactivity, cerebellar and cerebral neocortices had levels of the transporter >100-fold lower than the VMAT2-rich striatum and with no significant differences among the cortical regions. We conclude that human cerebellar and cerebral cortices contain negligible VMAT2 protein versus the striatum and, in this respect, all satisfy a criterion for a useful reference region for VMAT2 imaging. The slightly lower PET signal for VMAT2 binding in occipital (the currently preferred reference region) versus cerebellar cortex might not therefore be explained by differences in VMAT2 protein itself but possibly by other imaging variables, for example, partial volume effects.13Binding Potential Images

AXIAL SAGITAL CORONALAXIAL SAGITAL CORONAL

AXIAL SAGITAL CORONAL

Two tissue Compartment Model

Simplified Ref. Tissue Model

Simplified Ref. Tissue Model 2VERY NOISY!!!

14Analysis PipelineInter-frame Motion CorrectionNormalization:PET -> MR templateVoxel-wise group analysis: cocaine vs. control Voxel-wise2-tissueCompartmentModel

Voxel-wiseSRTM2Model

Voxel-wiseSRTMModel

FULL MODELSIMPLIFIED MODELSThis is less noisy, but constrained under the assumptionsone-tissue compartment fitting well with the reference (if not fitting well, it indicates the reference is not well chosen, it may have some specific binding potential) andthe delivery from plasma/vascular to the ROI and the reference is the same (R1)

Blood is only availble for 20 subjects.Voxel-wise could be very noisy, 2-T model needs to obtain 4 parameters, whereas simplified models only need to estimate 2 or 3 parameters, but constrained by the assumptions. They could be good b/c they are less noisy. 15NormalizationNormalizationStepsMethod 1PET -> PET templateMethod 2PET -> MR -> MR templateMethod 3PET -> MR -> MR template -> MNI spaceMethod 4PET -> MR -> MNI spaceWhy method 2?Our template is more similar to our images than MNI Subregions of striatum do not appear in standard atlases

ROIs manually drawn in the templateROIs transformed to the individual native spaceSRTM2: two-tissue for the ROI, one-tissue for the referenceThis is less noisy, but constrained under the assumptionsone-tissue compartment fitting well with the reference (if not fitting well, it indicates the reference is not well chosen, it may have some specific binding potential) andthe delivery from plasma/vascular to the ROI and the reference is the same (R1)16Normalization: DARTEL Template

Ashburner, Neuroimage 2007

24 MR images 1.5x1x1mm

ACPC reorientation

Tissue SegmentationTemplate generation

Deformation Maps

Composite template

The choice of reference region in positron emission tomography (PET) human brain imaging of the vesicular monoamine transporter 2 (VMAT2), a marker of striatal dopamine innervation, has been arbitrary, with cerebellar, whole cerebral, frontal, or occipital cortices used. To establish whether levels of VMAT2 are in fact low in these cortical areas, we measured VMAT2 protein distribution by quantitative immunoblotting in autopsied normal human brain (n=6). Four or five species of VMAT2 immunoreactivity (75, 55, 52, 45, 35kDa) were detected, which were all markedly reduced in intensity in nigrostriatal regions of patients with parkinsonian conditions versus matched controls (n=9 to 10 each). Using the intact VMAT2 immunoreactivity, cerebellar and cerebral neocortices had levels of the transporter >100-fold lower than the VMAT2-rich striatum and with no significant differences among the cortical regions. We conclude that human cerebellar and cerebral cortices contain negligible VMAT2 protein versus the striatum and, in this respect, all satisfy a criterion for a useful reference region for VMAT2 imaging. The slightly lower PET signal for VMAT2 binding in occipital (the currently preferred reference region) versus cerebellar cortex might not therefore be explained by differences in VMAT2 protein itself but possibly by other imaging variables, for example, partial volume effects.17Normalization: PET images

2) Apply PET -> MR estimation to BP images (NO RESLICE)

MR imageMutual Info N. Mutual Info Entropy Corr Coeff Norm. Cross Coeff1) Corregistration Reference frame PET to MR (ESTIMATE)3) Apply MR -> template deformation maps RESLICE (1.5x1.5x1.5mm)

The choice of reference region in positron emission tomography (PET) human brain imaging of the vesicular monoamine transporter 2 (VMAT2), a marker of striatal dopamine innervation, has been arbitrary, with cerebellar, whole cerebral, frontal, or occipital cortices used. To establish whether levels of VMAT2 are in fact low in these cortical areas, we measured VMAT2 protein distribution by quantitative immunoblotting in autopsied normal human brain (n=6). Four or five species of VMAT2 immunoreactivity (75, 55, 52, 45, 35kDa) were detected, which were all markedly reduced in intensity in nigrostriatal regions of patients with parkinsonian conditions versus matched controls (n=9 to 10 each). Using the intact VMAT2 immunoreactivity, cerebellar and cerebral neocortices had levels of the transporter >100-fold lower than the VMAT2-rich striatum and with no significant differences among the cortical regions. We conclude that human cerebellar and cerebral cortices contain negligible VMAT2 protein versus the striatum and, in this respect, all satisfy a criterion for a useful reference region for VMAT2 imaging. The slightly lower PET signal for VMAT2 binding in occipital (the currently preferred reference region) versus cerebellar cortex might not therefore be explained by differences in VMAT2 protein itself but possibly by other imaging variables, for example, partial volume effects.18Analysis PipelineInter-frame Motion CorrectionNormalization:PET -> MR templateVoxel-wise group analysis: cocaine vs. control Voxel-wise2-tissueCompartmentModel

Voxel-wiseSRTM2Model

Voxel-wiseSRTMModel

FULL MODELSIMPLIFIED MODELSThis is less noisy, but constrained under the assumptionsone-tissue compartment fitting well with the reference (if not fitting well, it indicates the reference is not well chosen, it may have some specific binding potential) andthe delivery from plasma/vascular to the ROI and the reference is the same (R1)

Blood is only availble for 20 subjects.Voxel-wise could be very noisy, 2-T model needs to obtain 4 parameters, whereas simplified models only need to estimate 2 or 3 parameters, but constrained by the assumptions. They could be good b/c they are less noisy. 19Statistical results: cocaine < control

SRTMSRTM26mm Gaussian smooth, mask BP>0.5, FWE corrected p100-fold lower than the VMAT2-rich striatum and with no significant differences among the cortical regions. We conclude that human cerebellar and cerebral cortices contain negligible VMAT2 protein versus the striatum and, in this respect, all satisfy a criterion for a useful reference region for VMAT2 imaging. The slightly lower PET signal for VMAT2 binding in occipital (the currently preferred reference region) versus cerebellar cortex might not therefore be explained by differences in VMAT2 protein itself but possibly by other imaging variables, for example, partial volume effects.20Comparative with ROI approach

Functional subdivisionAnatomical subdivisionpFWEcorrectedAssociate striatum0.002Dorsal caudate0.002Anterior Putamen0.007Caudate0.02Sensorimotor striatum0.002Posterior putamen 0.002Limbic striatum0.05Ventral striatum0.05Functional subdivisionAnatomical subdivisionCluster pFWEcorrectedN. cluster voxelsAssociate striatumDorsal caudate0.02566Anterior Putamen0.06611Sensorimotor striatumPosterior putamen 0.04430VOXEL WISE RESULTSPREVIOUS ROI RESULTSThe choice of reference region in positron emission tomography (PET) human brain imaging of the vesicular monoamine transporter 2 (VMAT2), a marker of striatal dopamine innervation, has been arbitrary, with cerebellar, whole cerebral, frontal, or occipital cortices used. To establish whether levels of VMAT2 are in fact low in these cortical areas, we measured VMAT2 protein distribution by quantitative immunoblotting in autopsied normal human brain (n=6). Four or five species of VMAT2 immunoreactivity (75, 55, 52, 45, 35kDa) were detected, which were all markedly reduced in intensity in nigrostriatal regions of patients with parkinsonian conditions versus matched controls (n=9 to 10 each). Using the intact VMAT2 immunoreactivity, cerebellar and cerebral neocortices had levels of the transporter >100-fold lower than the VMAT2-rich striatum and with no significant differences among the cortical regions. We conclude that human cerebellar and cerebral cortices contain negligible VMAT2 protein versus the striatum and, in this respect, all satisfy a criterion for a useful reference region for VMAT2 imaging. The slightly lower PET signal for VMAT2 binding in occipital (the currently preferred reference region) versus cerebellar cortex might not therefore be explained by differences in VMAT2 protein itself but possibly by other imaging variables, for example, partial volume effects.21Take home message

VOXEL WISE VS. ROI

BEST RESULTS WITH SRTM2 IN OUR PROJECT

CAUTION WITH SIMPLIFIED MODELS ASSUMPTIONS

MR TEMPLATE:

IMPROVES NORMALIZATION AND STATISTICSONLY DRAW ROIs IN TEMPLATE

The choice of reference region in positron emission tomography (PET) human brain imaging of the vesicular monoamine transporter 2 (VMAT2), a marker of striatal dopamine innervation, has been arbitrary, with cerebellar, whole cerebral, frontal, or occipital cortices used. To establish whether levels of VMAT2 are in fact low in these cortical areas, we measured VMAT2 protein distribution by quantitative immunoblotting in autopsied normal human brain (n=6). Four or five species of VMAT2 immunoreactivity (75, 55, 52, 45, 35kDa) were detected, which were all markedly reduced in intensity in nigrostriatal regions of patients with parkinsonian conditions versus matched controls (n=9 to 10 each). Using the intact VMAT2 immunoreactivity, cerebellar and cerebral neocortices had levels of the transporter >100-fold lower than the VMAT2-rich striatum and with no significant differences among the cortical regions. We conclude that human cerebellar and cerebral cortices contain negligible VMAT2 protein versus the striatum and, in this respect, all satisfy a criterion for a useful reference region for VMAT2 imaging. The slightly lower PET signal for VMAT2 binding in occipital (the currently preferred reference region) versus cerebellar cortex might not therefore be explained by differences in VMAT2 protein itself but possibly by other imaging variables, for example, partial volume effects.22THANK YOU FOR YOUR ATTENTION

REFERENCESNarendran et al. Am J Psychiatry 169: 55-63, 2012Narendran et al. Synapse 2011Boileau et al. J Neurosci: 98509856,2008Narendran et al. JPET vol. 333 no. 2: 533-539, 2010Wu and Carson, J Cereb Blood Flow & Metab: 14401452, 2002Lammertsma and Hume, Neuroimage 1996Ashburner, Neuroimage 2007Woods, J Comp Assit Tomography 1998

ACKNOWLEDGEMENTSCharles Laymon Seong-Gi KimRajesh Narendran Bill EddyJulie PriceMichael L HimesCarl BeckerScott MasonJames Ruszkiewicz Cristy MatanMatthew OborskiChris CieplyDavneet Minhas

23SMOOTHING

24NormalizationNormalizationStepsMethod 1PET -> PET templateMethod 2PET -> MR -> MR templateMethod 3PET -> MR -> MR template -> MNI spaceMethod 4PET -> MR -> MNI spaceWhy method 2?Our template is more similar to our images than MNI Subregions of striatum do not appear in standard atlasesEvaluation of semi-automatic procedure to draw the ROIs

ROIs manually drawn in the templateROIs transformed to the individual native spaceSRTM2: two-tissue for the ROI, one-tissue for the referenceThis is less noisy, but constrained under the assumptionsone-tissue compartment fitting well with the reference (if not fitting well, it indicates the reference is not well chosen, it may have some specific binding potential) andthe delivery from plasma/vascular to the ROI and the reference is the same (R1)25