FALCON for Motion Correction in PET Imaging - Your Tracer Experiences?

[LalithShiyam]

Hey GitHub enthusiasts! Have you explored the incredible potential of FALCON for motion correction in PET imaging studies using various tracers? Or have you gone beyond and utilized it for other modalities as well? Share your experiences and let's geek out together!

Kindly follow the following template:

• Modality: PT or MR? If PT, please indicate the tracer (e.g. FDG PET)
• Type of motion correction: Rigid or deformable?
• Motion area that was corrected: Brain, cardiac, whole-body?
• Your rating for FALCON: Average, Good, Excellent.
• Comments: Your thoughts/feedback on falcon!

[zax0s]

Hi

Thank you for making this tool openly available.

I recently gave it a go with whole-body PET dynamic data of a novel C-11 glyburide tracer (from a mult-bed step-and-shoot acquisition protocol).
Originally there was complex arm motion as well as hip movement between frames. After MoCo (deformable; in lazy-mode) all those motion patterns are not visible anymore.

As for the experience, the tool was simple to install (Ubuntu based system) and to use on my nifti data.

[LalithShiyam]

Hi @zax0s glad you find it useful - is it OK for you to share the entire tracer name?

[zax0s]

Sure, it was [11C] glyburide.

Marie S, Breuil L, Chalampalakis Z, Becquemont L, Verstuyft C, Lecoq AL, Caillé F, Gervais P, Lebon V, Comtat C, Bottlaender M, Tournier N. [11C]glyburide PET imaging for quantitative determination of the importance of Organic Anion-Transporting Polypeptide transporter function in the human liver and whole-body. Biomed Pharmacother. 2022 Dec;156:113994. doi: 10.1016/j.biopha.2022.113994. Epub 2022 Nov 9. PMID: 36411655.

[LalithShiyam]

Thank you and have fun with falcon!

[Abbaun]

Dear @LalithShiyam

I am currently trying to implement Falcon in my work flow with multiple PET tracers, and in my testing I have noticed something that puzzles me. My scan files are in the single nifti format (.nii) and I have noticed that after motion correction, the output file of falconz is twice as big as the original file. Working with multiple files of +650 MB (before motion correction) on +40 subjects becomes quite a lot of extra storage space used when motion correcting the files.
When I check the image, there are no additional frames and the dimension of the image are the same, so I wonder where all that extra information comes from?

Modality: PT, testing with [18F]-DOPA (dopamine tracer). I plan to use falcon on [11C]-donepezil (cholinergic tracer) and [11C]-(R)-PK11195 (inflammation tracer)
Type of motion correction: Rigid
Motion area that was corrected: Brain
Your rating for FALCON: Excellent (so far)

[LalithShiyam]

Hi @Abbaun, could be the data type of the written images. Can you quickly open your before MoCo image and tell me the dtype and do the same for after MoCo image?

Also thanks for the kind words and glad you like falcon! 🥳

[Abbaun]

Thanks for the quick response! It seems both files are float64.

[LalithShiyam]

Hi @Abbaun i will get back to you over the weekend. Was a busy week...

[LalithShiyam]

Hi @Abbaun,

I think the size bump is due to the floating point precision, by default it does double precision.
Is it possible to share one nifti file? for me to test it? And come up with possible solutions?

Cheers,
Lalith

[Abbaun]

Hi Lalith
I will try look into sharing a nifti file with you (just want to make sure I am not breaking any GDPR rules).

In regards to the precision. I am assuming you mean the precision (number of decimals) of the voxel values of the images? What is the reasoning for doubling the precision?
I would think that the output data could never have higher (meaningful) precision than the input. If anything, processing the data would reduce the precision as we apply some form of interpolation in the realignment of the images.

Best wishes,
Andreas