BigStitcher for multimodal registration/stitching



Does anyone have any experience/comments on using the BigStitcher plugin for multimodal registration and stitching? The end goal is quantitative image analysis by grabbing data from overlapping regions. BigStitcher was recommended to me as a good tool for stitching intra-modality, but if I understand correctly it may be capable of doing multimodal as well, and might be easier than integrating external registration tools

I’m quite new to the plugin and I have just begun playing around with it for regular intra-modal stitching, so I would appreciate any advice or suggestions about this topic. I’m also trying to automate the process through scripting.

For reference, here’s a graphic of what I’m trying to do. Boxes are the images I’m taking. The colored circles are fiducials for registration. Top left is the lowest resolution method (~50 micron voxels). The orange/green/blue boxes are singular tiles. Middle and left are the same resolution (~1 micron) and use the same optics, so should be aligned just by reading stage position. The black and red boxes are regular grids of many tiles.

Th original idea was registering left to middle, and by doing so registering it to the right and getting the sample data (transitively).



Hey @mpinkert,

I’ve used Bigstitcher successfully for intra-modality stitching, but have not tried inter-modality.

Short answer: try it, its great.

If you’re looking for mostly automatic stitching, then the main advantage of BigStitcher over other stuff is its ability to easily correct / manually align your tiles if things go wrong.

(Since, to my understanding, BigStitcher’s biggest improvements are with regard to the pipelining and interactivity of stitching more than the underlying automatic algorithms).

The documentation is quite good, so that’s a good place to start. But of course feel free to post back here for any specific issues.



Hi John,

Thanks for the tip on stitching correction - it led me to explore with that some more, and I do like how easy it is to correct if things go wrong. I think I’ve got a grasp on the basics of stitching, but now I’m diving into how the registration works. In particular, whether it would work at all for multimodal images.

@StephanPreibisch, does your lab have any thoughts or experience on using BigStitcher to register multimodal images, given approximate starting locations?

My naive understanding of the plugin says that this should be possible, given appropriate assignment of views/angles/coordinates, but I’m aware registration can be very finicky at times, and what works for microscopy tiles may not be good for microscopy <-> ultrasound (for example).