Batch Conversion of Cellomics C01 Files Slow

[Cammyron]

Hi All,

We are running a large screen ~120 384 well plates with 20ish fields of view per well through a Cellomics system. As the cellomics software is a bit lacking we are using custom code (written in MetaMoorph) to analyse all the images. To be able to do this we need to convert all the cellomics c01 files to tiff first.

We have a script written up in ImageJ that does the batch conversion using the LOCI tools, but we are hitting a few snags in regards to speed.

1. If everythign is ran locally the best we can acheive is around 20 images a minute. There are around 20,000 images per plate so this is a little bit too slow. The delay seems to be in reading the C01 header information. Is there anyway to speed this up?

2. If we run the script on files stored on the network (1GBps solid network connection) then it goes really slow. ABoout 3-4 images a minute.

Any ideas on how to speed this up would be greatly appreciated.

[Cammyron]

Ok, we have just done a copuple more tests. If there are not too many files in the folder the batch goes really fast. but when there are many >10,000 files in the folder it goes very slow.

If you manually open one of the files from a directory with >10,000 files it takes a while too. It seems to be indexing the folder contents everytime you open an image. Is there anyway to disable this?

[mlinkert]

You should see a significant performance improvement by using the "Open files individually" option. In an ImageJ macro, that would look like:

Code: Select all

run("Bio-Formats Importer", "open=/path/to/file.c01 open_files");


That makes sure that only the chosen file is opened - Bio-Formats will then not attempt to look for other files that belong to the same plate.

-Melissa

[Cammyron]

Hi Melisa,

Thanks for the response. The sript i am using already implements that code

Code: Select all

run("Bio-Formats Importer","open=["+dir1+list[i]+"] open_files");

The current plan of attack is to write something up in python to spilt the images into directories of say 2000 images, process them all and then shove it all back together.