Hi everyone, I’m a student in NeuroPoly Lab at Polytechnique Montreal. We are looking into BIDS for histology and would like to start a conversation with the community. Here is an example of the structure we propose for handling histology data. Comments and feedback are welcome!
Best,
Marie-Helene
1 Like
Welcome to NeuroStars and thank you for your interest in starting a conversation around extending BIDS into histology!
Perhaps posting over in the bids-specification GitHub repository issues section to get the conversation going - this area is more trafficked by our standardization community.
I found a post from a previously interested colleague asking about BIDS histology - may be beneficial to reach out.
Thank you,
Franklin
1 Like
This looks like a great start! I like the list of modalities, which looks consistent with how BIDS approaches it.
From a quick look, here are some thoughts:
-
You’ll probably also want to consider any metadata relevant to the interpretation of an image. For instance, perhaps certain microscope vendors, models or settings would be worth recording.
-
For histology, stains and dyes might be worth promoting into the filename to help distinguish things. In MRI, we have
ceagent(contrast-enhancing agent), which might be close enough semantically to be reused here, for example,sub-01_ceagent-nissl_OM.tiff. -
I’m not really up on my histology, so this may be a dumb suggestion, but if there’s only one way to image a given stain or dye, it’s possible that it would make more sense to promote these to the level of modality, instead of SEM/TEM/OM/CARS.
-
Organ or structure might be worth promoting into the filename as well.
-
In any event, I would strongly suggest settling on a single file format. The advantage of BIDS is that by constraining what data is BIDS, the more downstream tools can depend on it and spend more code on processing and less on handling disparate inputs. I would advocate a lossless format. For the sake of random-access, image compression algorithms may not be ideal. TIFF would probably be my suggestion.
-
You can also consider something like NIfTI, which permits 2-dimensional images and includes an affine matrix that can be used to provide information like the pixel size, direction of each dimension in anatomical coordinates. This would obviate your
PixelSizeandFovmetadata, and NIfTI can encode RGB values as well as scalars. The obvious downside is that histological tooling may not support NIfTI.
I agree with Franklin that this would be a good discussion to open on the GitHub issues to get further guidance on integrating with BIDS. It would also be good to try to reach out to other labs doing histology or labs/companies working on analytical software. Broad input (and commitment to implement) is critical for designing a standard that people use.
cc @crocodoyle
1 Like
Hi @franklin,
Thank you for your feedback, I will move this discussion to github issues here as you suggested.
Best,
Marie-Helene
Hi @effigies,
Thank you for your feedback and suggestions, it is much appreciated.
I’ve added to the example some metadata fields (manufacturer, software, institution) shared with other modalities.
I’ll take a closer look into specific settings, dyes vs modality, organs and file formats.
As you and Franklin suggested, I will move this discussion to github issues here.
Thanks again,
Marie-Helene
Hi @effigies,
Even though I’ve opened a github issue, it seems to make more sense to address your points here.
-
For stains and dyes, I’ve changed the metadata field “SamplePreparation” for both “SampleFixation” and “SampleStaining”, which are more specifics. I’m not sure at this point if the staining needs to be distinguishable on a filename basis but I’ll keep that in mind for future discussion. The same goes for organ.
-
As for the modalities (SEM/TEM/OM/CARS), we think they are essential to identify without ambiguities the images as they are commonly used to distinguish between different histology imaging techniques.
-
For the file formats, we agree that restricting to one file formats could be beneficial down the line. We are thinking about PNG which is lossless and smaller than TIFF. In addition, PNG would be compatible with other existing tools for dealing with histology images (compare to Nifti files).
Again, thank you for your feedback.
Feel free to add comments or suggestions directly on the doc here.
Best,
Marie-Helene