clarifications and changes for the point source acquisitions

[KrisThielemans]

@gschramm please check (and improve!)

[gschramm]

attached an svg of the point source positions (according to my understanding). if correct, can be included in the protocol as well.

I suppose what you want is 2 acq. where each time 2 point sources are in the FOV? or do you need 4 point sources in the FOV for both acqs?

[KrisThielemans]

Nice picture! In the text box on LPS axes, I'd prefer to specifically state direction (e.g. x = R->L, towards patient left)

@nkarakatsanis @AlanOsys please check. Maybe you wanted to the point sources "above" the centre. ATM they're below, which is practically harder I guess.

I suppose what you want is 2 acq. where each time 2 point sources are in the FOV? or do you need 4 point sources in the FOV for both acqs?

No, 4 acq., each with one point source, explicitly stated

PETSIRDTestData/doc/AcquisitionProtocols.md

Line 77 in aa980f3

- 1 source per acquisition

[AlanOsys]

The pictures look great, it’s very clean. Should we add the 3D printed phantom we will use to them? To give a clearer picture of what it should look like. With directions to which holes the tubes should be in.

[gschramm]

The pictures look great, it’s very clean. Should we add the 3D printed phantom we will use to them? To give a clearer picture of what it should look like. With directions to which holes the tubes should be in.

Note sure if everybody will use a 3D printed phantom holder for that ...

[gschramm]

Here is an updated version of the figure

[KrisThielemans]

Not sure if everybody will use a 3D printed phantom holder for that ...

agreed. I think if @nkarakatsanis wants to ship his holder around, he can provide the picture.

(I'd personally prefer to have z+3 a bit larger anyway, certainly for scanners without bed, and i seem to recall that this limitation came from the printed holder).

[KrisThielemans]

I suggest we modify/reduce the number of point source scans. Rationale:

• we need at least 3 locations to be sure about coordinate systems.
• if we put the 3rd source close to the edge of the FOV, we need bed movement really only checking translation. For this we need only 1 source (not 3).
• if we ask people to put point sources in different locations and move the bed, making sure that the bed is always in the same positions for each point source could be non-trivial
• @nkarakatsanis makes a good point about size of scanner. we will have some preclinical systems.
• vendor recon of 1 source with bed movement will show only 1 source, so we cannot use it for checking amount of bed movement. However, we can do an "internal" check: detected shift in position of source has to be equal to bed translation
• I'd keep both CBM and step-and-shoot, as handling the former is harder.

This all leads to:

• stationary bed: 3 point sources. positions (0, -1, L/2 -1), (1,0,0), (0,-(R/2),0), with L the axial FOV length and R the FOV radius.
• step-and-shoot bed movement for last point source, moving over min L/5, max L/3
• CBM (if supported) for last point source over same range (could be in reverse direction, or restart from original position)

@gschramm @AlanOsys @nkarakatsanis @pjmark. ok?

Note: we need to suggest something for duration of CBM scan.

[gschramm]

• I agree that 1 acq. with 3 point sources is easier, takes lesss time and is less prone to errors.
• ideally this measurement could be also used to verify PET vs CT, or PET vs MR alignment if the point sources are visible in CT/MR
• for the the step and shoot check, I would simply recommend to keep the 3 sources on the table and two acquire a 2 bed pos scan.

[gschramm]

@KrisThielemans isn't it "safer" to require 4 point source, because 3 points will always for 1 plane?

[KrisThielemans]

• I agree that 1 acq. with 3 point sources is easier, takes lesss time and is less prone to errors.

I'm afraid we had discussed this, and agreed to 1 ptsrc per acq. So, I'm proposing 3 acquisition here...

• ideally this measurement could be also used to verify PET vs CT, or PET vs MR alignment if the point sources are visible in CT/MR

yes, but often they're not (unless you have a specifically defined phantom, such as the GE VQC phantom for the Signa...)

• for the the step and shoot check, I would simply recommend to keep the 3 sources on the table and two acquire a 2 bed pos scan.

yeah, but we have only 1 point source...

[KrisThielemans]

@KrisThielemans isn't it "safer" to require 4 point source, because 3 points will always for 1 plane?

of course. More acquisitions are fine for me :-). We could say "at least 3"...

[gschramm]

@KrisThielemans isn't it "safer" to require 4 point source, because 3 points will always for 1 plane?

of course. More acquisitions are fine for me :-). We could say "at least 3"...

I agree with less acq. is better. But timewise there is no difference between an acq with 3 or 4 source in the FOV.

[KrisThielemans]

Given our previous experience with debugging, we strongly suggest to have a single source in every acquisition. (Of course, once it works, there is no reason for it!).

Point source acquisitions hopefully shouldn't take long, just make them very active (roughly equivalent to activity of 4 point sources :-)

[nkarakatsanis]

I agree with "at least 3" point sources and 3 separate static acquisitions, i.e. 1 acquisition per source. That is we have 3 acquisitions with a stationary bed.

Why to set the second radial position to R/2? This could result in a not-so-easy-to-position radial location. Of course it would be approximate but then we will not be able to use any printed phantom with predefined radial locations. I therefore suggest we use a fixed absolute position. R=1cm and R=12cm sound like a reasonable option, unless you wish to suggest a different radial offset. I therefore suggest the following three locations: (0, -1, L/2 -1), (1,0,0), (0,-12,0), with L the axial FOV length

Then we need 1 additional acquisition for 1 source (the one placed at (1,0,0) at the first bed position) for the second bed position in step-and-shoot mode.

We also need 1 addditional acquisition for the same source in CBM mode. The CBM mode can extend up to 1/3rd of the AFOV. I suggest a CBM speed that allows to cover the scanner AFOV in 2min. That would imply slower CBM speed for shorter AFOV scanners and faster CBM speeed for long AFOV scanners. 2min/bed is usually the fastest we scan with modern short AFOV clinical PET systems. Do you agree?

[pjmark]

Hi Kris - happy with the proposal. I can work with any data, and from my experience the best tests I get is from chunky phantoms scanned for a really long time, which we may do later on.

For now, quick question here - what does -1 or 1 means? Sorry if silly question (early flight this morning...)

• stationary bed: 3 point sources. positions (0, -1, L/2 -1), (1,0,0), (0,-(R/2),0), with L the axial FOV length and R the FOV radius.

[pjmark]

Hi Nicolas: R=12 cm might only work for some scanners, likely not the brain-only, if you mean radius and if we care at all at this stage.

Why to set the second radial position to R/2? This could result in a not-so-easy-to-position radial location. Of course it would be approximate but then we will not be able to use any printed phantom with predefined radial locations. I therefore suggest we use a fixed absolute position. R=1cm and R=12cm sound like a reasonable option, unless you wish to suggest a different radial offset. I therefore suggest the following three locations: (0, -1, L/2 -1), (1,0,0), (0,-12,0), with L the axial FOV length

[pjmark]

BTW, are we scared to put points in the centre? :D Sometimes funny things happen there...

[nkarakatsanis]

Hi @pjmark, I recommend to avoid the transaxial center and use R=1cm radial offset instead. NEMA prtocols suggest the same.

Regarding R=12cm, it would work for brain scanners (i do not know any brain PET scanner with a transaxial FOV ofd less than 25cm diameter. Typically we are between 28 and 40cm. But yes it would not work for a prelinical system. Maybe we can state that "for scanenrs with diameter less than 28cm we recommend as the second radial offset an integer multiple of 1 mm such that the radial distance between that point and the transaxial edge is at least 2 mm".

[KrisThielemans]

I agree with "at least 3" point sources and 3 separate static acquisitions, i.e. 1 acquisition per source. That is we have 3 acquisitions with a stationary bed.

great

Why to set the second radial position to R/2? This could result in a not-so-easy-to-position radial location. Of course it would be approximate but then we will not be able to use any printed phantom with predefined radial locations. I therefore suggest we use a fixed absolute position. R=1cm and R=12cm sound like a reasonable option, unless you wish to suggest a different radial offset. I therefore suggest the following three locations: (0, -1, L/2 -1), (1,0,0), (0,-12,0), with L the axial FOV length

apologies, I meant to be able to cover preclinical (or brain) scanners, as @pjmark says, but obviously for a whole body scanner R/2 is pretty inconvenient. We can say "12 cm, or for small FOV scanners R/2" (or something like you said, although that's bit more complicated)

Then we need 1 additional acquisition for 1 source (the one placed at (1,0,0) at the first bed position) for the second bed position in step-and-shoot mode.

I prefer the (0,-12,0) as more off-centre, but I don't mind.

We also need 1 addditional acquisition for the same source in CBM mode. The CBM mode can extend up to 1/3rd of the AFOV. I suggest a CBM speed that allows to cover the scanner AFOV in 2min. That would imply slower CBM speed for shorter AFOV scanners and faster CBM speeed for long AFOV scanners. 2min/bed is usually the fastest we scan with modern short AFOV clinical PET systems. Do you agree?

ok for me!

[nkarakatsanis]

Why to set the second radial position to R/2? This could result in a not-so-easy-to-position radial location. Of course it would be approximate but then we will not be able to use any printed phantom with predefined radial locations. I therefore suggest we use a fixed absolute position. R=1cm and R=12cm sound like a reasonable option, unless you wish to suggest a different radial offset. I therefore suggest the following three locations: (0, -1, L/2 -1), (1,0,0), (0,-12,0), with L the axial FOV length

apologies, I meant to be able to cover preclinical (or brain) scanners, as @pjmark says, but obviously for a whole body scanner R/2 is pretty inconvenient. We can say "12 cm, or for small FOV scanners R/2" (or something like you said, although that's bit more complicated)

I prefer the simplicity of absolute integer values rather than relative ones. I do agree that this may make the choice of those integer values more complicated but at the end i found calculations easier than experiments, so I vote for radial offsets that are integer multiplies of mm's .

Then we need 1 additional acquisition for 1 source (the one placed at (1,0,0) at the first bed position) for the second bed position in step-and-shoot mode.

I prefer the (0,-12,0) as more off-centre, but I don't mind.

OK for me to use the (0,-12,0) for either the 2nd bed position or the CBM acquisition.

[KrisThielemans]

I've come round to @gschramm 's comment that we prefer 4 non-planar sources. Then everything is unique.

[KrisThielemans]

@nkarakatsanis you wrote

I prefer the simplicity of absolute integer values rather than relative ones. I do agree that this may make the choice of those integer values more complicated but at the end i found calculations easier than experiments, so I vote for radial offsets that are integer multiplies of mm's .

We say "approximate positions", and currently all others are in cm (typo in the text!). If we start to talk about mm positions, then positioning has to be very accurate, which I think most people will not achieve anyway. So, I'll stick to my phrasing for now...

[nkarakatsanis]

@KrisThielemans you wrote

We say "approximate positions", and currently all others are in cm (typo in the text!). If we start to talk about mm positions, then positioning has to be very accurate, which I think most people will not achieve anyway. So, I'll stick to my phrasing for now...

I think mm accuracy is doable with the use of a 3D printed phantom where predetrmined integer multiples of mm positions are engraved. However i agree that for those that do not wish to use such a phantom and also for axial bed translations it may not be easy to achieve 1mm positioning accuracy. Therefore I am fine with the use of cm units and for the use of the term "approximate positioning". Nevertheless, I still think it is easier to use integer multiples of mm (then converted to cm) for the radial and axial offsets.

[KrisThielemans]

@nkarakatsanis @gschramm @AlanOsys @rgwells2023 @pjmark I've incorporated all of the above comments now. Obviously, @gschramm 's figure isn't included as it will need to be updated.

I'd like to merge this Monday morning, such that we can send it to the ETSI list.

We can always do another PR for extra modifications (including the figure).

[gschramm]

@nkarakatsanis @gschramm @AlanOsys @rgwells2023 @pjmark I've incorporated all of the above comments now. Obviously, @gschramm 's figure isn't included as it will need to be updated.

I'd like to merge this Monday morning, such that we can send it to the ETSI list.

We can always do another PR for extra modifications (including the figure).

I will update the figure as soon as the new positions are final (takes 10min).

[KrisThielemans]

I'm going to merge this now (we can keep moving our point sources :-)).

@gschramm, please do a different PR with the figure, feel free to wait till tomorrow in case somebody has major objections.