We are currently testing OMERO at our research institute. It works well for now, but we have a problem about the uploading of very big data (from light sheet microscopy, in particular). Raw data is generated as multiple tif files (1 tif file for each channel/each Z-depth), but our scientists need to analyze them on Imaris, so they need to convert the tif files to an .ims file. And to make things more easy, they want to be able to upload their .ims files on OMERO. Here comes the problems.
When trying to import the original dataset of tif files (864 files x around 10-11 Mb each), it only takes almost 12 minutes (when only selecting the first file: the other ones are concatenated to form a stack on OMERO). The other phases (“Reading Pixels”, “Generating Thumbnails”…) are negligible, almost several seconds.
When importing the .ims file (around 4 Gb) on OMERO.insight, the import takes almost 3 minutes. But the “Reading Pixels” step is extremely long, almost 1 hour.
I would like to know how can it be explained, and how can i reduce the time taken by this “Reading Pixels” step, because it is the real bottleneck. One of the scientists i work with suggested me to “suppress” the thumbnail creation step (if there’s a way to do it), but i’m not so sure that it will solve the problem: when i examine the file structure on imageJ and when i compare it with the way the file is stored on the OMERO server, i think the problem is more linked with the way the .ims file format stores the data.
When i open it on ImageJ/Fiji:
When i observe the result of the import on OMERO.web (still pending, not all the series are yet imported, the importation is still at the “Reading Pixels” step):
I can see that the way it is stored onOMERO mirrors the way it is read on ImageJ/Fiji. Maybe it’s a part of the explanation.
EDIT: It just finished now:
I can see that even not all the series (5 on 6) have been uploaded.
Best regards, Marc.