[000350] Missing voxels spatial information

[mcelotto]

I’m Marco Celotto, one of the participants in the NeuroDataRehack 2025 event. As part of our hackathon project, my team is reanalyzing the rich dataset 000350 from the Ahrens lab.
One issue we’re facing is mapping single-cell ROIs to specific brain regions (e.g., LC or NE-MO, as defined in Mu et al., Cell, 2019). Our plan is to use a zebrafish brain atlas to map voxels from the volumetric imaging data into anatomical regions, and then use the ROI masks stored in /processing/ophys/VolumeSegmentation/NeuronVolumeSegmentation/voxel_mask to assign individual neurons to those regions.
To do this, we need to know how voxels map to physical space (in micrometers). We couldn't find metadata on the x-y pixel size or the spacing between z-planes in the NWB files. Speaking with Misha Ahrens today, he told us that each x-y pixel corresponds to 0.40325 µm, and that the distance between z-planes varies between 5 µm and 8 µm across sessions.
From our understanding of the methods in Mu et al., this spatial information was likely used to assign ROIs to regions like L-MO. We’re wondering whether the z-spacing used in each session is stored somewhere in the dataset or could be recovered, to allow accurate voxel-to-region mapping. @bendichter suggested we open this issue and mentioned that @CodyCBakerPhD might be able to help retrieve this information. We completely understand if it’s not recoverable - if that's the case, we’ll adjust our analysis plan accordingly. Thanks for your time!

[CodyCBakerPhD]

Hi @mcelotto

Unfortunately, after checking the original source data, I do not see any records of the exact microscope metadata, including the depth settings (or other values related to grid spacing or field of view)

I only see some .log files containing information regarding the array shapes and temporal frequency of the sampling; see ahrens-lab-to-nwb/issues/1 for a thorough discussion of what they look like

Perhaps there might be some way to indirectly estimate it given the shape/width of the outer larvel zebrafish profile as seen in the microscopy (top relative to bottom)? If it is super important

[mcelotto]

Hi @CodyCBakerPhD,

Thanks so much for your prompt reply - it's very helpful to know that the depth settings are likely not available. I found this tool for aligning zebrafish volumetric imaging data to reference atlases, but it seems to rely on knowing voxel dimensions in micrometers to optimize the registration.

Mapping cells onto an atlas would be very useful for both the whole-brain information transfer analyses we’re planning and for interpreting our results more broadly. I suppose it might still be possible to adapt the alignment tools by trying a range of z-spacing values, but that likely goes beyond what we can achieve during the hackathon. Hopefully, we’ll be able to revisit this after the event and explore it further.

Thanks again for your help!

[bendichter]

Perhaps there might be some way to indirectly estimate it given the shape/width of the outer larvel zebrafish profile as seen in the microscopy (top relative to bottom)?

This sounds promising. You could find the z-length of the zebrafish in the atlas and set the z planar distance to match the subject's z to the atlas's z. Even if it is not completely correct due to variability in size of individuals, it should allow you to register the subject to the atlas.