Experience of multi-view imaging and data registration

Hi all,

I was wondering what SPIM users are using to register their multi-view datasets captured at different angles. If using the Multiview Reconstruction plugin in FIJI how easy do you find this, do you have any general advice about how to make it more likely registration is successful?

We’re finding that our sample changes position considerably when it is rotated so quite big changes in xyz via the sample stage are required and can even mean there is no xyz coordinate overlap to bring the sample back to the same focus once rotated, this means we have to manually try to overlay the datasets in the Multiview reconstruction plugin before even starting to detect interest points. How much of this can be alleviated through accurate sample mounting e.g. the sample is close to the centre of rotation of the rotation motor by being in the middle of whatever vessel is used FEP tubing etc.

Thanks,

Matt

Hi Matt,

I have been moving the xy stage to get the same parts of the sample in the FOV with rotations (takes a bit of set-up), since you must have the same detections in the different angles for registration/fusion. You can also try smaller rotation angles (but will take more memory/longer processing), adding more beads (if you don’t have enough corresponding detection for registration), or centering the sample in the vessel (like you said). Manually moving the components and using combination of all 4 algorithms in the plugin also improves the chances. Hope this helps

  • Brian
1 Like

Ok thanks Brian, I’m not entirely sure what the problem is, our beads are very bright in respect to the sample and i read that its better if they’re weak but i think that’s more for sample visualisation rather than it benefitting the registration algorithms… Also i suspected the large shifts of the stage required after each rotation meaning there was no overlap between the different stacks was perhaps hampering it. I think we should try smaller angles rather than just 0 and 180.

Thanks,

Matt