I’m new to the MRTrix. I encountered a PC spec issue.
Here are the questions for you:
What kind of PC specs are required to visualize 100 million virtual white matter fibers with mrview? (Visualization of 50,000 virtual white matter fibers is possible.)
What role does the GPU play in mrtrix?
We are attempting to visualize virtual white matter fibers from GPe (globus pallidus external) to STN (subthalamic nucleus) using the following procedure.
Step1. use tckgen to output 100 million virtual white matter fibers in the whole brain to a tck file⇨Done. Successfully.
Step2. use tckedit to output virtual white matter fibers from GPe to STN to a tck file based on the tck file created in Step1⇨Done. Successfully.
Step3. I will visualize the tck file created in Step2 with mrview. ⇨Failed.
I realized that mrview was freezed when visualizing the tck file.
The DTI used in tckgen was imaged under the following conditions.
・HARDI
・b=2000
・slice thickness 2mm
Our spec is as follows:
OS: Ubuntu22.04 (WSL2)
CPU: Intel Xeon E7-8890 v3 18core x 4 (total 72core)
RAM: 512 GB
I would appreciate your suggestions, feedback, and recommendations.
When visualising within mrview, quite a lot! The streamlines data is sent to GPU memory, and needs to fit there to be rendered successfully. 100 million streamlines will require ~40GB of storage – far too large to fit into all but the most expensive GPUs. On my system, I can render a 4GB tractogram with 10 million streamlines, but it takes a while to load, and the frame rate isn’t fantastic…
More to the point: there is rarely any good reason to display such a large tractogram, there are far too many streamlines to really appreciate the detail. What I would typically recommend for quick visual inspection is to pick out a subset of the streamlines (maybe the first 100,000) using tckedit and display that, e.g.:
tckedit tracks_10m.tck -num 100k tracks_100k.tck
I also note you don’t show any detail about the GPU in the specs for your system, so I can’t really comment on that aspect.
Are you sure your GPU is a Matrox G200…? That card dates back to 1998! That said, Wikipedia does note that:
Matrox G200 series, especially the G200e is still a popular choice for server motherboard manufacturers, like Dell’s PowerEdge series, due to its robustness, low power consumption and limited features needed just for VGA display.
The issue really is that you’re using a server that isn’t designed for display or GPU workloads. But as far as I can tell, it should be possible to insert a better card in there – you don’t need anything too fancy, but a mid-range nvidia GeForce would already allow you to speed up eddy and (hence dwifslpreproc) substantially, and improve your display performance enormously…