# Diffusion tractography performance limit questions

**URL:** <https://discourse.slicer.org/t/diffusion-tractography-performance-limit-questions/2073>\
**Category:** SlicerDMRI\
**Tags:** diffusion\
**Created:** [February 12, 2018, 6:17pm UTC](https://discourse.slicer.org/t/diffusion-tractography-performance-limit-questions/2073 "2018-02-12T18:17:37Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![jamesjcook](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.slicer.org/jamesjcook/32/1177_2.png) [@jamesjcook](https://discourse.slicer.org/u/jamesjcook)\
**Post date:** [February 12, 2018, 6:17pm UTC](https://discourse.slicer.org/t/diffusion-tractography-performance-limit-questions/2073/1 "2018-02-12T18:17:37Z")

</div>

I’m trying to use slicer to visualize tracks. My diffusion volumes were 275x225x450 with 0.2mm voxels.

Has anyone explored the limits of track visualization? Memory usage appears strangely high when I load in my tracks, the vtk fiberbundles are less than 1GiB on disk, but use 25+GiB in RAM.

My total number of tracks may be high, it seems to indicate I’m loading 200,000+ tracks in about 20 different vtk fiberbundle files. Some of which have 45,000, others having only 200-1000.

When these tracks are initially loaded the interface is frustratingly slow/laggy, even after disabling the display modules. I’ve tested in slicer 4.6, 4.8, and nightly of 2018-02-06. After saving and loading all tracks and the scene, and switching visualization to lines colored by segment, instead of tubes, the interface behaves reasonably.

---

<div class="post-metadata">

**Author:** ![ihnorton](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.slicer.org/ihnorton/32/9_2.png) [@ihnorton](https://discourse.slicer.org/u/ihnorton)\
**Post date:** [February 13, 2018, 2:46pm UTC](https://discourse.slicer.org/t/diffusion-tractography-performance-limit-questions/2073/2 "2018-02-13T14:46:57Z")

</div>

I’m working on performance improvements for this issue. My current largest test-set is about 400MB, so if you are able to share your dataset, I would be interested in looking at it (via dropbox, gdrive, etc. – if you want to email me: `inorton at bwh dot harvard dot edu`).

> switching visualization to lines colored by segment, instead of tubes

The performance difference between these two should go away soon, thanks to the VTK opengl2 improvements.