Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-14 Thread Greve, Douglas N.,Ph.D. 
 >>But in the case where --psf 0 --no-tfe are specified, isn't the high 
resolution of gtmseg lost when rescaling intensities to mean reference 
region ?
Yes


On 2/14/19 5:25 PM, Matthieu Vanhoutte wrote:
>
> External Email - Use Caution
>
>
>
> Le jeu. 14 févr. 2019 à 21:36, Greve, Douglas N.,Ph.D. 
> mailto:dgr...@mgh.harvard.edu>> a écrit :
>
>
>
> On 2/14/19 2:25 PM, Matthieu Vanhoutte wrote:
> >          External Email - Use Caution
> >
> >> Le 14 févr. 2019 à 19:37, Greve, Douglas N.,Ph.D.
> mailto:dgr...@mgh.harvard.edu>> a écrit :
> >>
> >>
> >>
> >> On 2/14/19 1:14 PM, Matthieu Vanhoutte wrote:
> >>>          External Email - Use Caution
> >>>
> >>>
> >>>
> >>> Le jeu. 14 févr. 2019 à 18:31, Greve, Douglas N.,Ph.D.
> >>> mailto:dgr...@mgh.harvard.edu>
> >> a
> écrit :
> >>>
> >>>
> >>>
> >>>     On 2/14/19 12:16 PM, Matthieu VANHOUTTE wrote:
>             External Email - Use Caution
> 
>  Hi Douglas,
> 
>  So to register PET to gtmseg space mri_gtmpvc command use nearest
>  neighbor, interpolation right ?
> >>>     It does not map the PET to the gtmseg space. It maps the
> gtmseg to
> >>>     the
> >>>     PET space. Since this is a segmentation, it uses nearest
> neighbor.
> >>>     But
> >>>     the algorithm is much more complicated; see the paper.
> >>>
> >>>
> >>> Thank you I will look into the paper, I was wondering how not
> to loose
> >>> the gtmseg resolution ?
> >> I thought you wanted to turn off all PVC? If you want to
> correct for the
> >> tissue fraction effect, then don't use --no-tfe
> > Oh, I had to miss something. I didn’t know PVC had several
> components, whose one is tissue fraction effect… What are the
> different corrections applied during PVC ?
> Tissue fraction effect and correction for blur induced by the low res
> nature of pet. See the paper.
> > When papers mention no PVC they principally don’t correct for
> tissue fraction effect ?
> I don't know. Generally, people don't give many details on their
> method.
> >
> > I wanted to know what is the interest of computing a high
> resolution segmentation (gtmseg) if it goes to PET space and
> resliced to its lower resolution ?
> As I said before, this is not what happens. PETsurfer takes the
> TFE into
> account.
> Please see the paper for how it is done.
>
>
> But in the case where --psf 0 --no-tfe are specified, isn't the high 
> resolution of gtmseg lost when rescaling intensities to mean reference 
> region ?
>
> >
> >>> Concerning RBV PVC in which order and what type of
> interpolation are
> >>> used since PVE corrected output is on gtmseg_space ?
> >> This question does not make sense in the context of RBV. Each
> ROI gets a
> >> value, and this value is then assigned to all the voxels in
> that ROI in
> >> gtmseg prior to apply the RBV algorithm.
> > Ok thank you, so in RBV case native PET is interpolated
> (trilinear?) first to gtmseg space ?
> >
>  And what about the interpolation method to use when
> projecting PVE
>  corrected PET on surface  ?
> >>>     I usually use nearest neighbor to avoid interpolation
> >>>
> >>>
> >>> Should you precise this on PETSurfer wiki since on the
> mri_vol2surf
> >>> command trilinear interpolation is used by default ?
> >> By default, mri_vol2surf uses nearest neighbor. mri_vol2vol
> uses trilin
> >> by default. An unfortunate inconsistency I made about 15 years
> ago and
> >> have been too afraid to change it:)
> > Thank you I didn’t know this one and supposed the trilinear as
> default as it is done in moi_vol2vol. Good to know it :)
> >
> >>> Best.
> >>>
>  Thanks,
> 
>  Matthieu
> 
>  On 14/02/2019 17:44, Greve, Douglas N.,Ph.D. wrote:
> > The GTM operates on regions, so you can't do trilin interp. The
> > operations that you describe below are not used in mri_gtmpvc.
> >>>     If  you
> > are trying to get something close to what the GTM does, then
> >>>     use nearest
> > neighbor
> >
> > On 2/14/19 3:43 AM, Matthieu Vanhoutte wrote:
> >>              External Email - Use Caution
> >>
> >> Hi Douglas,
> >>
> >> Could you help me concerning my previous questions ?
> >>
> >> Thanks,
> >> Matthieu
> >>
> >>> Le 11 févr. 2019 à 17:37, Matthieu VANHOUTTE
> >>>      
>  >>
> >>>     a écrit :
> >>> Hi Douglas,
>

Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-14 Thread Matthieu Vanhoutte 
External Email - Use Caution

Le jeu. 14 févr. 2019 à 21:36, Greve, Douglas N.,Ph.D. <
dgr...@mgh.harvard.edu> a écrit :

>
>
> On 2/14/19 2:25 PM, Matthieu Vanhoutte wrote:
> >  External Email - Use Caution
> >
> >> Le 14 févr. 2019 à 19:37, Greve, Douglas N.,Ph.D. <
> dgr...@mgh.harvard.edu> a écrit :
> >>
> >>
> >>
> >> On 2/14/19 1:14 PM, Matthieu Vanhoutte wrote:
> >>>  External Email - Use Caution
> >>>
> >>>
> >>>
> >>> Le jeu. 14 févr. 2019 à 18:31, Greve, Douglas N.,Ph.D.
> >>> mailto:dgr...@mgh.harvard.edu>> a écrit :
> >>>
> >>>
> >>>
> >>> On 2/14/19 12:16 PM, Matthieu VANHOUTTE wrote:
> External Email - Use Caution
> 
>  Hi Douglas,
> 
>  So to register PET to gtmseg space mri_gtmpvc command use nearest
>  neighbor, interpolation right ?
> >>> It does not map the PET to the gtmseg space. It maps the gtmseg to
> >>> the
> >>> PET space. Since this is a segmentation, it uses nearest neighbor.
> >>> But
> >>> the algorithm is much more complicated; see the paper.
> >>>
> >>>
> >>> Thank you I will look into the paper, I was wondering how not to loose
> >>> the gtmseg resolution ?
> >> I thought you wanted to turn off all PVC? If you want to correct for the
> >> tissue fraction effect, then don't use --no-tfe
> > Oh, I had to miss something. I didn’t know PVC had several components,
> whose one is tissue fraction effect… What are the different corrections
> applied during PVC ?
> Tissue fraction effect and correction for blur induced by the low res
> nature of pet. See the paper.
> > When papers mention no PVC they principally don’t correct for tissue
> fraction effect ?
> I don't know. Generally, people don't give many details on their method.
> >
> > I wanted to know what is the interest of computing a high resolution
> segmentation (gtmseg) if it goes to PET space and resliced to its lower
> resolution ?
> As I said before, this is not what happens. PETsurfer takes the TFE into
> account.
> Please see the paper for how it is done.
>

But in the case where --psf 0 --no-tfe are specified, isn't the high
resolution of gtmseg lost when rescaling intensities to mean reference
region ?

>
> >>> Concerning RBV PVC in which order and what type of interpolation are
> >>> used since PVE corrected output is on gtmseg_space ?
> >> This question does not make sense in the context of RBV. Each ROI gets a
> >> value, and this value is then assigned to all the voxels in that ROI in
> >> gtmseg prior to apply the RBV algorithm.
> > Ok thank you, so in RBV case native PET is interpolated (trilinear?)
> first to gtmseg space ?
> >
>  And what about the interpolation method to use when projecting PVE
>  corrected PET on surface  ?
> >>> I usually use nearest neighbor to avoid interpolation
> >>>
> >>>
> >>> Should you precise this on PETSurfer wiki since on the mri_vol2surf
> >>> command trilinear interpolation is used by default ?
> >> By default, mri_vol2surf uses nearest neighbor. mri_vol2vol uses trilin
> >> by default. An unfortunate inconsistency I made about 15 years ago and
> >> have been too afraid to change it:)
> > Thank you I didn’t know this one and supposed the trilinear as default
> as it is done in moi_vol2vol. Good to know it :)
> >
> >>> Best.
> >>>
>  Thanks,
> 
>  Matthieu
> 
>  On 14/02/2019 17:44, Greve, Douglas N.,Ph.D. wrote:
> > The GTM operates on regions, so you can't do trilin interp. The
> > operations that you describe below are not used in mri_gtmpvc.
> >>> If  you
> > are trying to get something close to what the GTM does, then
> >>> use nearest
> > neighbor
> >
> > On 2/14/19 3:43 AM, Matthieu Vanhoutte wrote:
> >>  External Email - Use Caution
> >>
> >> Hi Douglas,
> >>
> >> Could you help me concerning my previous questions ?
> >>
> >> Thanks,
> >> Matthieu
> >>
> >>> Le 11 févr. 2019 à 17:37, Matthieu VANHOUTTE
> >>> mailto:matthieuvanhou...@gmail.com>>
> >>> a écrit :
> >>> Hi Douglas,
> >>>
> >>> Thanks for clarification. Why use "nearest neighbor" instead
> >>> of classical "trilinear" interpolation ?
> >>> Should it be done as well when projecting from volume to
> >>> surface with mri_vol2surf ?
> >>> Best,
> >>>
> >>> Matthieu
> >>>
> >>> On 11/02/2019 17:32, Greve, Douglas N.,Ph.D. wrote:
>  It will not used trilinear interp. Try it with nearest neighbor
> 
>  On 2/10/19 11:55 AM, Matthieu Vanhoutte wrote:
> >   External Email - Use Caution
> >
> > Dear Douglas,
> >
> > Before using "mri_gtmpvc --psf 0 --no-tfe -- rbv" according
> >>> your advice to obtain voxel-wise output rescaled without PVC
> >>> correction, I have done sequentially the different steps to obtain
> >>> this output. However, when comparing voxel-wise rescaled

Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-14 Thread Greve, Douglas N.,Ph.D. 


On 2/14/19 2:25 PM, Matthieu Vanhoutte wrote:
>  External Email - Use Caution
>
>> Le 14 févr. 2019 à 19:37, Greve, Douglas N.,Ph.D.  a 
>> écrit :
>>
>>
>>
>> On 2/14/19 1:14 PM, Matthieu Vanhoutte wrote:
>>>  External Email - Use Caution
>>>
>>>
>>>
>>> Le jeu. 14 févr. 2019 à 18:31, Greve, Douglas N.,Ph.D.
>>> mailto:dgr...@mgh.harvard.edu>> a écrit :
>>>
>>>
>>>
>>> On 2/14/19 12:16 PM, Matthieu VANHOUTTE wrote:
External Email - Use Caution

 Hi Douglas,

 So to register PET to gtmseg space mri_gtmpvc command use nearest
 neighbor, interpolation right ?
>>> It does not map the PET to the gtmseg space. It maps the gtmseg to
>>> the
>>> PET space. Since this is a segmentation, it uses nearest neighbor.
>>> But
>>> the algorithm is much more complicated; see the paper.
>>>
>>>
>>> Thank you I will look into the paper, I was wondering how not to loose
>>> the gtmseg resolution ?
>> I thought you wanted to turn off all PVC? If you want to correct for the
>> tissue fraction effect, then don't use --no-tfe
> Oh, I had to miss something. I didn’t know PVC had several components, whose 
> one is tissue fraction effect… What are the different corrections applied 
> during PVC ?
Tissue fraction effect and correction for blur induced by the low res 
nature of pet. See the paper.
> When papers mention no PVC they principally don’t correct for tissue fraction 
> effect ?
I don't know. Generally, people don't give many details on their method.
>
> I wanted to know what is the interest of computing a high resolution 
> segmentation (gtmseg) if it goes to PET space and resliced to its lower 
> resolution ?
As I said before, this is not what happens. PETsurfer takes the TFE into 
account.
Please see the paper for how it is done.
>
>>> Concerning RBV PVC in which order and what type of interpolation are
>>> used since PVE corrected output is on gtmseg_space ?
>> This question does not make sense in the context of RBV. Each ROI gets a
>> value, and this value is then assigned to all the voxels in that ROI in
>> gtmseg prior to apply the RBV algorithm.
> Ok thank you, so in RBV case native PET is interpolated (trilinear?) first to 
> gtmseg space ?
>
 And what about the interpolation method to use when projecting PVE
 corrected PET on surface  ?
>>> I usually use nearest neighbor to avoid interpolation
>>>
>>>
>>> Should you precise this on PETSurfer wiki since on the mri_vol2surf
>>> command trilinear interpolation is used by default ?
>> By default, mri_vol2surf uses nearest neighbor. mri_vol2vol uses trilin
>> by default. An unfortunate inconsistency I made about 15 years ago and
>> have been too afraid to change it:)
> Thank you I didn’t know this one and supposed the trilinear as default as it 
> is done in moi_vol2vol. Good to know it :)
>
>>> Best.
>>>
 Thanks,

 Matthieu

 On 14/02/2019 17:44, Greve, Douglas N.,Ph.D. wrote:
> The GTM operates on regions, so you can't do trilin interp. The
> operations that you describe below are not used in mri_gtmpvc.
>>> If  you
> are trying to get something close to what the GTM does, then
>>> use nearest
> neighbor
>
> On 2/14/19 3:43 AM, Matthieu Vanhoutte wrote:
>>  External Email - Use Caution
>>
>> Hi Douglas,
>>
>> Could you help me concerning my previous questions ?
>>
>> Thanks,
>> Matthieu
>>
>>> Le 11 févr. 2019 à 17:37, Matthieu VANHOUTTE
>>> mailto:matthieuvanhou...@gmail.com>>
>>> a écrit :
>>> Hi Douglas,
>>>
>>> Thanks for clarification. Why use "nearest neighbor" instead
>>> of classical "trilinear" interpolation ?
>>> Should it be done as well when projecting from volume to
>>> surface with mri_vol2surf ?
>>> Best,
>>>
>>> Matthieu
>>>
>>> On 11/02/2019 17:32, Greve, Douglas N.,Ph.D. wrote:
 It will not used trilinear interp. Try it with nearest neighbor

 On 2/10/19 11:55 AM, Matthieu Vanhoutte wrote:
>   External Email - Use Caution
>
> Dear Douglas,
>
> Before using "mri_gtmpvc --psf 0 --no-tfe -- rbv" according
>>> your advice to obtain voxel-wise output rescaled without PVC
>>> correction, I have done sequentially the different steps to obtain
>>> this output. However, when comparing voxel-wise rescaled output
>>> (rbv.nii.gz) between the two methods, final values at each voxel
>>> are not the same…
> To obtain first rescaled voxel-wise output I have used
>>> sequentially these steps:
> 1.   Concatenate transforms from PET space to gtmseg.mgz
> 2.   Register native PET into gtmseg space (trilin.
>>> interpolation)
> 3.   Compute mean PET inside reference regions (with
>>> regions masks from gtmseg.mgz)
> 4.   Compute intensity normalized PET images in

Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-14 Thread Matthieu Vanhoutte 
External Email - Use Caution

> Le 14 févr. 2019 à 19:37, Greve, Douglas N.,Ph.D.  a 
> écrit :
> 
> 
> 
> On 2/14/19 1:14 PM, Matthieu Vanhoutte wrote:
>> 
>> External Email - Use Caution
>> 
>> 
>> 
>> Le jeu. 14 févr. 2019 à 18:31, Greve, Douglas N.,Ph.D. 
>> mailto:dgr...@mgh.harvard.edu>> a écrit :
>> 
>> 
>> 
>>On 2/14/19 12:16 PM, Matthieu VANHOUTTE wrote:
>>>   External Email - Use Caution
>>> 
>>> Hi Douglas,
>>> 
>>> So to register PET to gtmseg space mri_gtmpvc command use nearest
>>> neighbor, interpolation right ?
>>It does not map the PET to the gtmseg space. It maps the gtmseg to
>>the
>>PET space. Since this is a segmentation, it uses nearest neighbor.
>>But
>>the algorithm is much more complicated; see the paper.
>> 
>> 
>> Thank you I will look into the paper, I was wondering how not to loose 
>> the gtmseg resolution ?
> I thought you wanted to turn off all PVC? If you want to correct for the 
> tissue fraction effect, then don't use --no-tfe

Oh, I had to miss something. I didn’t know PVC had several components, whose 
one is tissue fraction effect… What are the different corrections applied 
during PVC ? When papers mention no PVC they principally don’t correct for 
tissue fraction effect ?

I wanted to know what is the interest of computing a high resolution 
segmentation (gtmseg) if it goes to PET space and resliced to its lower 
resolution ?

>> 
>> Concerning RBV PVC in which order and what type of interpolation are 
>> used since PVE corrected output is on gtmseg_space ?
> This question does not make sense in the context of RBV. Each ROI gets a 
> value, and this value is then assigned to all the voxels in that ROI in 
> gtmseg prior to apply the RBV algorithm.

Ok thank you, so in RBV case native PET is interpolated (trilinear?) first to 
gtmseg space ?

>> 
>>> 
>>> And what about the interpolation method to use when projecting PVE
>>> corrected PET on surface  ?
>>I usually use nearest neighbor to avoid interpolation
>> 
>> 
>> Should you precise this on PETSurfer wiki since on the mri_vol2surf 
>> command trilinear interpolation is used by default ?
> By default, mri_vol2surf uses nearest neighbor. mri_vol2vol uses trilin 
> by default. An unfortunate inconsistency I made about 15 years ago and 
> have been too afraid to change it:)

Thank you I didn’t know this one and supposed the trilinear as default as it is 
done in moi_vol2vol. Good to know it :)

>> 
>> Best.
>> 
>>> 
>>> Thanks,
>>> 
>>> Matthieu
>>> 
>>> On 14/02/2019 17:44, Greve, Douglas N.,Ph.D. wrote:
 The GTM operates on regions, so you can't do trilin interp. The
 operations that you describe below are not used in mri_gtmpvc.
>>If  you
 are trying to get something close to what the GTM does, then
>>use nearest
 neighbor
 
 On 2/14/19 3:43 AM, Matthieu Vanhoutte wrote:
> External Email - Use Caution
> 
> Hi Douglas,
> 
> Could you help me concerning my previous questions ?
> 
> Thanks,
> Matthieu
> 
>> Le 11 févr. 2019 à 17:37, Matthieu VANHOUTTE
>>mailto:matthieuvanhou...@gmail.com>>
>>a écrit :
>> 
>> Hi Douglas,
>> 
>> Thanks for clarification. Why use "nearest neighbor" instead
>>of classical "trilinear" interpolation ?
>> 
>> Should it be done as well when projecting from volume to
>>surface with mri_vol2surf ?
>> 
>> Best,
>> 
>> Matthieu
>> 
>> On 11/02/2019 17:32, Greve, Douglas N.,Ph.D. wrote:
>>> It will not used trilinear interp. Try it with nearest neighbor
>>> 
>>> On 2/10/19 11:55 AM, Matthieu Vanhoutte wrote:
  External Email - Use Caution
 
 Dear Douglas,
 
 Before using "mri_gtmpvc --psf 0 --no-tfe -- rbv" according
>>your advice to obtain voxel-wise output rescaled without PVC
>>correction, I have done sequentially the different steps to obtain
>>this output. However, when comparing voxel-wise rescaled output
>>(rbv.nii.gz) between the two methods, final values at each voxel
>>are not the same…
 
 To obtain first rescaled voxel-wise output I have used
>>sequentially these steps:
 1.   Concatenate transforms from PET space to gtmseg.mgz
 2.   Register native PET into gtmseg space (trilin.
>>interpolation)
 3.   Compute mean PET inside reference regions (with
>>regions masks from gtmseg.mgz)
 4.   Compute intensity normalized PET images in gtmseg
>>space
 
 Doesn’t the "mri_gtmpvc --psf 0 --no-tfe -- rdv" command
>>use these same steps ?
 
 Best,
 Matthieu
 
 ___
 Freesurfer mailing list
 Freesurfer@nmr.mgh.harvard.edu
>>

Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-14 Thread Greve, Douglas N.,Ph.D. 


On 2/14/19 1:14 PM, Matthieu Vanhoutte wrote:
>
> External Email - Use Caution
>
>
>
> Le jeu. 14 févr. 2019 à 18:31, Greve, Douglas N.,Ph.D. 
> mailto:dgr...@mgh.harvard.edu>> a écrit :
>
>
>
> On 2/14/19 12:16 PM, Matthieu VANHOUTTE wrote:
> >          External Email - Use Caution
> >
> > Hi Douglas,
> >
> > So to register PET to gtmseg space mri_gtmpvc command use nearest
> > neighbor, interpolation right ?
> It does not map the PET to the gtmseg space. It maps the gtmseg to
> the
> PET space. Since this is a segmentation, it uses nearest neighbor.
> But
> the algorithm is much more complicated; see the paper.
>
>
> Thank you I will look into the paper, I was wondering how not to loose 
> the gtmseg resolution ?
I thought you wanted to turn off all PVC? If you want to correct for the 
tissue fraction effect, then don't use --no-tfe
>
> Concerning RBV PVC in which order and what type of interpolation are 
> used since PVE corrected output is on gtmseg_space ?
This question does not make sense in the context of RBV. Each ROI gets a 
value, and this value is then assigned to all the voxels in that ROI in 
gtmseg prior to apply the RBV algorithm.
>
> >
> > And what about the interpolation method to use when projecting PVE
> > corrected PET on surface  ?
> I usually use nearest neighbor to avoid interpolation
>
>
> Should you precise this on PETSurfer wiki since on the mri_vol2surf 
> command trilinear interpolation is used by default ?
By default, mri_vol2surf uses nearest neighbor. mri_vol2vol uses trilin 
by default. An unfortunate inconsistency I made about 15 years ago and 
have been too afraid to change it:)
>
> Best.
>
> >
> > Thanks,
> >
> > Matthieu
> >
> > On 14/02/2019 17:44, Greve, Douglas N.,Ph.D. wrote:
> >> The GTM operates on regions, so you can't do trilin interp. The
> >> operations that you describe below are not used in mri_gtmpvc.
> If  you
> >> are trying to get something close to what the GTM does, then
> use nearest
> >> neighbor
> >>
> >> On 2/14/19 3:43 AM, Matthieu Vanhoutte wrote:
> >>>            External Email - Use Caution
> >>>
> >>> Hi Douglas,
> >>>
> >>> Could you help me concerning my previous questions ?
> >>>
> >>> Thanks,
> >>> Matthieu
> >>>
>  Le 11 févr. 2019 à 17:37, Matthieu VANHOUTTE
> mailto:matthieuvanhou...@gmail.com>>
> a écrit :
> 
>  Hi Douglas,
> 
>  Thanks for clarification. Why use "nearest neighbor" instead
> of classical "trilinear" interpolation ?
> 
>  Should it be done as well when projecting from volume to
> surface with mri_vol2surf ?
> 
>  Best,
> 
>  Matthieu
> 
>  On 11/02/2019 17:32, Greve, Douglas N.,Ph.D. wrote:
> > It will not used trilinear interp. Try it with nearest neighbor
> >
> > On 2/10/19 11:55 AM, Matthieu Vanhoutte wrote:
> >>             External Email - Use Caution
> >>
> >> Dear Douglas,
> >>
> >> Before using "mri_gtmpvc --psf 0 --no-tfe -- rbv" according
> your advice to obtain voxel-wise output rescaled without PVC
> correction, I have done sequentially the different steps to obtain
> this output. However, when comparing voxel-wise rescaled output
> (rbv.nii.gz) between the two methods, final values at each voxel
> are not the same…
> >>
> >> To obtain first rescaled voxel-wise output I have used
> sequentially these steps:
> >> 1.       Concatenate transforms from PET space to gtmseg.mgz
> >> 2.       Register native PET into gtmseg space (trilin.
> interpolation)
> >> 3.       Compute mean PET inside reference regions (with
> regions masks from gtmseg.mgz)
> >> 4.       Compute intensity normalized PET images in gtmseg
> space
> >>
> >> Doesn’t the "mri_gtmpvc --psf 0 --no-tfe -- rdv" command
> use these same steps ?
> >>
> >> Best,
> >> Matthieu
> >>
> >> ___
> >> Freesurfer mailing list
> >> Freesurfer@nmr.mgh.harvard.edu
> 
> >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> > ___
> > Freesurfer mailing list
> > Freesurfer@nmr.mgh.harvard.edu
> 
> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> >>> ___
> >>> Freesurfer mailing list
> >>> Freesurfer@nmr.mgh.harvard.edu
> 
> >>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> >>

Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-14 Thread Matthieu Vanhoutte 
External Email - Use Caution

Le jeu. 14 févr. 2019 à 18:31, Greve, Douglas N.,Ph.D. <
dgr...@mgh.harvard.edu> a écrit :

>
>
> On 2/14/19 12:16 PM, Matthieu VANHOUTTE wrote:
> >  External Email - Use Caution
> >
> > Hi Douglas,
> >
> > So to register PET to gtmseg space mri_gtmpvc command use nearest
> > neighbor, interpolation right ?
> It does not map the PET to the gtmseg space. It maps the gtmseg to the
> PET space. Since this is a segmentation, it uses nearest neighbor. But
> the algorithm is much more complicated; see the paper.
>

Thank you I will look into the paper, I was wondering how not to loose the
gtmseg resolution ?

Concerning RBV PVC in which order and what type of interpolation are used
since PVE corrected output is on gtmseg_space ?

>
> > And what about the interpolation method to use when projecting PVE
> > corrected PET on surface  ?
> I usually use nearest neighbor to avoid interpolation
>

Should you precise this on PETSurfer wiki since on the mri_vol2surf command
trilinear interpolation is used by default ?

Best.

>
> > Thanks,
> >
> > Matthieu
> >
> > On 14/02/2019 17:44, Greve, Douglas N.,Ph.D. wrote:
> >> The GTM operates on regions, so you can't do trilin interp. The
> >> operations that you describe below are not used in mri_gtmpvc. If  you
> >> are trying to get something close to what the GTM does, then use nearest
> >> neighbor
> >>
> >> On 2/14/19 3:43 AM, Matthieu Vanhoutte wrote:
> >>>External Email - Use Caution
> >>>
> >>> Hi Douglas,
> >>>
> >>> Could you help me concerning my previous questions ?
> >>>
> >>> Thanks,
> >>> Matthieu
> >>>
>  Le 11 févr. 2019 à 17:37, Matthieu VANHOUTTE <
> matthieuvanhou...@gmail.com> a écrit :
> 
>  Hi Douglas,
> 
>  Thanks for clarification. Why use "nearest neighbor" instead of
> classical "trilinear" interpolation ?
> 
>  Should it be done as well when projecting from volume to surface with
> mri_vol2surf ?
> 
>  Best,
> 
>  Matthieu
> 
>  On 11/02/2019 17:32, Greve, Douglas N.,Ph.D. wrote:
> > It will not used trilinear interp. Try it with nearest neighbor
> >
> > On 2/10/19 11:55 AM, Matthieu Vanhoutte wrote:
> >> External Email - Use Caution
> >>
> >> Dear Douglas,
> >>
> >> Before using "mri_gtmpvc --psf 0 --no-tfe -- rbv" according your
> advice to obtain voxel-wise output rescaled without PVC correction, I have
> done sequentially the different steps to obtain this output. However, when
> comparing voxel-wise rescaled output (rbv.nii.gz) between the two methods,
> final values at each voxel are not the same…
> >>
> >> To obtain first rescaled voxel-wise output I have used sequentially
> these steps:
> >> 1.   Concatenate transforms from PET space to gtmseg.mgz
> >> 2.   Register native PET into gtmseg space (trilin.
> interpolation)
> >> 3.   Compute mean PET inside reference regions (with regions
> masks from gtmseg.mgz)
> >> 4.   Compute intensity normalized PET images in gtmseg space
> >>
> >> Doesn’t the "mri_gtmpvc --psf 0 --no-tfe -- rdv" command use these
> same steps ?
> >>
> >> Best,
> >> Matthieu
> >>
> >> ___
> >> Freesurfer mailing list
> >> Freesurfer@nmr.mgh.harvard.edu
> >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> > ___
> > Freesurfer mailing list
> > Freesurfer@nmr.mgh.harvard.edu
> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> >>> ___
> >>> Freesurfer mailing list
> >>> Freesurfer@nmr.mgh.harvard.edu
> >>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> >> ___
> >> Freesurfer mailing list
> >> Freesurfer@nmr.mgh.harvard.edu
> >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> > ___
> > Freesurfer mailing list
> > Freesurfer@nmr.mgh.harvard.edu
> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
>
>
> ___
> Freesurfer mailing list
> Freesurfer@nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-14 Thread Greve, Douglas N.,Ph.D. 


On 2/14/19 12:16 PM, Matthieu VANHOUTTE wrote:
>  External Email - Use Caution
>
> Hi Douglas,
>
> So to register PET to gtmseg space mri_gtmpvc command use nearest
> neighbor, interpolation right ?
It does not map the PET to the gtmseg space. It maps the gtmseg to the 
PET space. Since this is a segmentation, it uses nearest neighbor. But 
the algorithm is much more complicated; see the paper.
>
> And what about the interpolation method to use when projecting PVE
> corrected PET on surface  ?
I usually use nearest neighbor to avoid interpolation
>
> Thanks,
>
> Matthieu
>
> On 14/02/2019 17:44, Greve, Douglas N.,Ph.D. wrote:
>> The GTM operates on regions, so you can't do trilin interp. The
>> operations that you describe below are not used in mri_gtmpvc. If  you
>> are trying to get something close to what the GTM does, then use nearest
>> neighbor
>>
>> On 2/14/19 3:43 AM, Matthieu Vanhoutte wrote:
>>>External Email - Use Caution
>>>
>>> Hi Douglas,
>>>
>>> Could you help me concerning my previous questions ?
>>>
>>> Thanks,
>>> Matthieu
>>>
 Le 11 févr. 2019 à 17:37, Matthieu VANHOUTTE  
 a écrit :

 Hi Douglas,

 Thanks for clarification. Why use "nearest neighbor" instead of classical 
 "trilinear" interpolation ?

 Should it be done as well when projecting from volume to surface with 
 mri_vol2surf ?

 Best,

 Matthieu

 On 11/02/2019 17:32, Greve, Douglas N.,Ph.D. wrote:
> It will not used trilinear interp. Try it with nearest neighbor
>
> On 2/10/19 11:55 AM, Matthieu Vanhoutte wrote:
>> External Email - Use Caution
>>
>> Dear Douglas,
>>
>> Before using "mri_gtmpvc --psf 0 --no-tfe -- rbv" according your advice 
>> to obtain voxel-wise output rescaled without PVC correction, I have done 
>> sequentially the different steps to obtain this output. However, when 
>> comparing voxel-wise rescaled output (rbv.nii.gz) between the two 
>> methods, final values at each voxel are not the same…
>>
>> To obtain first rescaled voxel-wise output I have used sequentially 
>> these steps:
>> 1.   Concatenate transforms from PET space to gtmseg.mgz
>> 2.   Register native PET into gtmseg space (trilin. interpolation)
>> 3.   Compute mean PET inside reference regions (with regions masks 
>> from gtmseg.mgz)
>> 4.   Compute intensity normalized PET images in gtmseg space
>>
>> Doesn’t the "mri_gtmpvc --psf 0 --no-tfe -- rdv" command use these same 
>> steps ?
>>
>> Best,
>> Matthieu
>>
>> ___
>> Freesurfer mailing list
>> Freesurfer@nmr.mgh.harvard.edu
>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> ___
> Freesurfer mailing list
> Freesurfer@nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
>>> ___
>>> Freesurfer mailing list
>>> Freesurfer@nmr.mgh.harvard.edu
>>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
>> ___
>> Freesurfer mailing list
>> Freesurfer@nmr.mgh.harvard.edu
>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
> ___
> Freesurfer mailing list
> Freesurfer@nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer


___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-14 Thread Matthieu VANHOUTTE 
External Email - Use Caution

Hi Douglas,

So to register PET to gtmseg space mri_gtmpvc command use nearest 
neighbor, interpolation right ?

And what about the interpolation method to use when projecting PVE 
corrected PET on surface  ?

Thanks,

Matthieu

On 14/02/2019 17:44, Greve, Douglas N.,Ph.D. wrote:
> The GTM operates on regions, so you can't do trilin interp. The
> operations that you describe below are not used in mri_gtmpvc. If  you
> are trying to get something close to what the GTM does, then use nearest
> neighbor
>
> On 2/14/19 3:43 AM, Matthieu Vanhoutte wrote:
>>   External Email - Use Caution
>>
>> Hi Douglas,
>>
>> Could you help me concerning my previous questions ?
>>
>> Thanks,
>> Matthieu
>>
>>> Le 11 févr. 2019 à 17:37, Matthieu VANHOUTTE  
>>> a écrit :
>>>
>>> Hi Douglas,
>>>
>>> Thanks for clarification. Why use "nearest neighbor" instead of classical 
>>> "trilinear" interpolation ?
>>>
>>> Should it be done as well when projecting from volume to surface with 
>>> mri_vol2surf ?
>>>
>>> Best,
>>>
>>> Matthieu
>>>
>>> On 11/02/2019 17:32, Greve, Douglas N.,Ph.D. wrote:
 It will not used trilinear interp. Try it with nearest neighbor

 On 2/10/19 11:55 AM, Matthieu Vanhoutte wrote:
>External Email - Use Caution
>
> Dear Douglas,
>
> Before using "mri_gtmpvc --psf 0 --no-tfe -- rbv" according your advice 
> to obtain voxel-wise output rescaled without PVC correction, I have done 
> sequentially the different steps to obtain this output. However, when 
> comparing voxel-wise rescaled output (rbv.nii.gz) between the two 
> methods, final values at each voxel are not the same…
>
> To obtain first rescaled voxel-wise output I have used sequentially these 
> steps:
> 1.Concatenate transforms from PET space to gtmseg.mgz
> 2.Register native PET into gtmseg space (trilin. interpolation)
> 3.Compute mean PET inside reference regions (with regions masks 
> from gtmseg.mgz)
> 4.Compute intensity normalized PET images in gtmseg space
>
> Doesn’t the "mri_gtmpvc --psf 0 --no-tfe -- rdv" command use these same 
> steps ?
>
> Best,
> Matthieu
>
> ___
> Freesurfer mailing list
> Freesurfer@nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
 ___
 Freesurfer mailing list
 Freesurfer@nmr.mgh.harvard.edu
 https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
>> ___
>> Freesurfer mailing list
>> Freesurfer@nmr.mgh.harvard.edu
>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
>
> ___
> Freesurfer mailing list
> Freesurfer@nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-14 Thread Greve, Douglas N.,Ph.D. 
The GTM operates on regions, so you can't do trilin interp. The 
operations that you describe below are not used in mri_gtmpvc. If  you 
are trying to get something close to what the GTM does, then use nearest 
neighbor

On 2/14/19 3:43 AM, Matthieu Vanhoutte wrote:
>  External Email - Use Caution
>
> Hi Douglas,
>
> Could you help me concerning my previous questions ?
>
> Thanks,
> Matthieu
>
>> Le 11 févr. 2019 à 17:37, Matthieu VANHOUTTE  a 
>> écrit :
>>
>> Hi Douglas,
>>
>> Thanks for clarification. Why use "nearest neighbor" instead of classical 
>> "trilinear" interpolation ?
>>
>> Should it be done as well when projecting from volume to surface with 
>> mri_vol2surf ?
>>
>> Best,
>>
>> Matthieu
>>
>> On 11/02/2019 17:32, Greve, Douglas N.,Ph.D. wrote:
>>> It will not used trilinear interp. Try it with nearest neighbor
>>>
>>> On 2/10/19 11:55 AM, Matthieu Vanhoutte wrote:
   External Email - Use Caution

 Dear Douglas,

 Before using "mri_gtmpvc --psf 0 --no-tfe -- rbv" according your advice to 
 obtain voxel-wise output rescaled without PVC correction, I have done 
 sequentially the different steps to obtain this output. However, when 
 comparing voxel-wise rescaled output (rbv.nii.gz) between the two methods, 
 final values at each voxel are not the same…

 To obtain first rescaled voxel-wise output I have used sequentially these 
 steps:
 1. Concatenate transforms from PET space to gtmseg.mgz
 2. Register native PET into gtmseg space (trilin. interpolation)
 3. Compute mean PET inside reference regions (with regions masks from 
 gtmseg.mgz)
 4. Compute intensity normalized PET images in gtmseg space

 Doesn’t the "mri_gtmpvc --psf 0 --no-tfe -- rdv" command use these same 
 steps ?

 Best,
 Matthieu

 ___
 Freesurfer mailing list
 Freesurfer@nmr.mgh.harvard.edu
 https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
>>> ___
>>> Freesurfer mailing list
>>> Freesurfer@nmr.mgh.harvard.edu
>>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
>
> ___
> Freesurfer mailing list
> Freesurfer@nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer


___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-14 Thread Matthieu Vanhoutte 
External Email - Use Caution

Hi Douglas,

Could you help me concerning my previous questions ?

Thanks,
Matthieu

> Le 11 févr. 2019 à 17:37, Matthieu VANHOUTTE  a 
> écrit :
> 
> Hi Douglas,
> 
> Thanks for clarification. Why use "nearest neighbor" instead of classical 
> "trilinear" interpolation ?
> 
> Should it be done as well when projecting from volume to surface with 
> mri_vol2surf ?
> 
> Best,
> 
> Matthieu
> 
> On 11/02/2019 17:32, Greve, Douglas N.,Ph.D. wrote:
>> It will not used trilinear interp. Try it with nearest neighbor
>> 
>> On 2/10/19 11:55 AM, Matthieu Vanhoutte wrote:
>>>  External Email - Use Caution
>>> 
>>> Dear Douglas,
>>> 
>>> Before using "mri_gtmpvc --psf 0 --no-tfe -- rbv" according your advice to 
>>> obtain voxel-wise output rescaled without PVC correction, I have done 
>>> sequentially the different steps to obtain this output. However, when 
>>> comparing voxel-wise rescaled output (rbv.nii.gz) between the two methods, 
>>> final values at each voxel are not the same…
>>> 
>>> To obtain first rescaled voxel-wise output I have used sequentially these 
>>> steps:
>>> 1.  Concatenate transforms from PET space to gtmseg.mgz
>>> 2.  Register native PET into gtmseg space (trilin. interpolation)
>>> 3.  Compute mean PET inside reference regions (with regions masks from 
>>> gtmseg.mgz)
>>> 4.  Compute intensity normalized PET images in gtmseg space
>>> 
>>> Doesn’t the "mri_gtmpvc --psf 0 --no-tfe -- rdv" command use these same 
>>> steps ?
>>> 
>>> Best,
>>> Matthieu
>>> 
>>> ___
>>> Freesurfer mailing list
>>> Freesurfer@nmr.mgh.harvard.edu
>>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
>> 
>> ___
>> Freesurfer mailing list
>> Freesurfer@nmr.mgh.harvard.edu
>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer


___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Re: [Freesurfer] PETSurfer: "mri_gtmpvc --psf 0 --no-tfe" voxel-wise output

2019-02-11 Thread Greve, Douglas N.,Ph.D. 
It will not used trilinear interp. Try it with nearest neighbor

On 2/10/19 11:55 AM, Matthieu Vanhoutte wrote:
>  External Email - Use Caution
>
> Dear Douglas,
>
> Before using "mri_gtmpvc --psf 0 --no-tfe -- rbv" according your advice to 
> obtain voxel-wise output rescaled without PVC correction, I have done 
> sequentially the different steps to obtain this output. However, when 
> comparing voxel-wise rescaled output (rbv.nii.gz) between the two methods, 
> final values at each voxel are not the same…
>
> To obtain first rescaled voxel-wise output I have used sequentially these 
> steps:
> 1.Concatenate transforms from PET space to gtmseg.mgz
> 2.Register native PET into gtmseg space (trilin. interpolation)
> 3.Compute mean PET inside reference regions (with regions masks from 
> gtmseg.mgz)
> 4.Compute intensity normalized PET images in gtmseg space
>
> Doesn’t the "mri_gtmpvc --psf 0 --no-tfe -- rdv" command use these same steps 
> ?
>
> Best,
> Matthieu
>
> ___
> Freesurfer mailing list
> Freesurfer@nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer


___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer