Tuning for large or raw images

Hello! Not sure if it’s about new features or an existing one. We have a microscopy system that returns huge files. Cellprofiler on an average Mac with 4G RAM gets stuck reading even a part of an image (e.g. the part is 4096 x 4096 pixels x 3 colors). Question: is it possible to tune somehow Cellprofiler to read large images or raw data? If it reads raw, then we would partially process the source file and feed Cellprofiler with the result. Or maybe there are some other way to handle huge sets?

Thanks in advance.


Unless you’re running a 64-bit machine, CP will not be able to handle images of the size you describe. We strongly suggest that you consider downsampling the image prior to loading in CellProfiler (e.g., image much larger than 2000 x 2000 pixels). Alternately, you might try splitting the image into non-overlapping tiles and processing each separately.


It was 64-bit Mac and max size of non-stuck images was 4096x3072. Sometimes splitting doesn’t work good, e.g. when cells have to be measured (a cell can be divided between different tiles) and downsampling of a 12000x8000 source image to 2000x2000 doesn’t make sense.
But anyway thanks for your response.

Try a 64-bit Windows machine, if possible. Unfortunately, CP for OSX is not really 64-bit (regardless of the processor architecture), and though we’re aware of the problem and working on it, it is really an underlying issue with some python libraries that we use (I believe it is wx in this case). But 64-bit Windows runs a true 64-bit CP.

Another simple help for memory issues is to split up your color images into multiple grayscale images before running in CP (using e.g. ImageJ).