Tracking Algorithm method choice

[Michaël Bernier]

Hi, again, awesome software, thanks for that!

I'm using DSI studio on Windows, and there is a small bug in the interface; I can't resize the "Step T3c" column showing the parameters, thus for some of them I can only guess what it is ;). I tried to look in the documentation, but can't find "Voxel". Is it FACT?

For example, 'Tracking Algorithm', there are 3 options: "Voxel...ing", "Stremline..er)" and "RK4".

With my very unusual dataset (I'm doing vascular tractography, don't interpret too much the figure), only "Voxel...ing" seems to work with my data (following the anatomy), with variable results at different step sizes. With "Streamline", I get very "straight" tracts when it works. If my guess is right, that "Voxel" is in fact "FACT", it should be very similar to "Streamline", no?

[Frank Yeh]

Wow. this picture is amazing.

"Voxel tracking" is an entirely new algorithm I developed several years ago, but I did not have time to write up a paper for it.

Are you interested in a collaboration? You can be the first author and your supervisor the corresponding author. I will provide a write-up for the "voxel" tracking algorithm.

As for the windows layout, DSI Studio appears very differently on my side, and I will see if I can figure out the problem.

--
You received this message because you are subscribed to the Google Groups "DSI Studio" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/69faf539-50f7-4af0-b941-93032ac397f8n%40googlegroups.com.

[Adam Raikes]

Hi Frank,

Following the thread for my own curiosity here. What exactly is the voxel tracking doing vs streamline?

Mouse brain images attached below for the sake of cool pictures. Streamline tracking first followed by voxel tracking

[Frank Yeh]

Voxel tracking is a voxel-wise clustering method, and there is no "tracking"

It uses voxels as the unit to link nearby voxels together as a path. The track is recreated from the linked voxels.

The idea is very different from conventional streamline fiber tracking.

Frank

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/2ea344d8-a019-42f2-a841-3262dc782e06n%40googlegroups.com.

[Michaël Bernier]

@Frank 
Thanks for the precisions! In fact, I will present my work at ISMRM, so I guess I'll have to talk a little bit about your method. Quick question: what role plays the "step" parameter in 'voxel tracking'? Is it the distance it looks for clustering the ODFs in order to decide where to go?
And of course I would be interested in a collaboration! In fact, I already started a paper about those vascular tracts, and your method could fit there (I also acquire 'traditional' dMRI).

@Adam

Nice! In my case, I get extreme versions of your images; if I use 'streamline', the tracts lose all relation with the anatomy, but for 'voxel tracking', although it looks more curly, it follows the anatomy very well. I'll try to generate a comparison later, it's.. impressive.

[Frank Yeh]

Thanks for the precisions! In fact, I will present my work at ISMRM, so I guess I'll have to talk a little bit about your method. Quick question: what role plays the "step" parameter in 'voxel tracking'? Is it the distance it looks for clustering the ODFs in order to decide where to go?

No, it is not used in voxel tracking.

 

And of course I would be interested in a collaboration! In fact, I already started a paper about those vascular tracts, and your method could fit there (I also acquire 'traditional' dMRI).

Sounds great. Look forward to hearing further from you.

Frank

 

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/c8bcf7e7-021f-49fe-9cf8-190404f43c0an%40googlegroups.com.

[Michaël Bernier]

Hum 'step size' does have an impact. For example, a small step doesn't work at all with some of my data, while a large step (5 to 10) make it like the picture I shared previously.

[Frank Yeh]

My mistake. I am checking the code and will get back to you.

Frank

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/5e0b6412-6498-432a-8866-cd21e8023705n%40googlegroups.com.

[Frank Yeh]

It seems that the step size is used to convert the minimum length to the number of steps, but this won't work correctly in voxel tracking because the step size is fixed.

Consequently, increasing the step size here will cause the  minimum length to malfunction and remove most tracks.

If you relax the minimum length to zero, it should give the same result.

I will fix this bug in the next release.

Frank