Question about proposed use

[SamWallace]

Hi

I've been following this software for a while now whilst I've been doing other parts of my research but I haven't set aside enough time to fully assess the capabilities of it.

I have a fairly quick question about my particular intended use.

I have a large set of images that generally contain only one whole cell. My current method of analysis has been to determine a set of ROIs for each cell, then perform some kind of image processing to pull out data about the histograms, then export that data into an excel spreadsheet to create a crude database.

What I'd like to be able to do is to only define the ROIs for each cell once, then have the software automatically run a given set of image processing commands and pull out the data and build the excel spreadsheet, or database.

Currently if I wish to change some part of my analysis it requires manually processing every image again. Even with scripts and macros to speed it up, it still takes a considerable amount of time.

So my question is. Can the OMERO server and clients be used in such a fashion that once I've imported all my images, I can then have a list of 'cells' which correspond to a set of ROIs for a given image. Then can I have a routine that performs the same analysis steps on each 'cell' and outputs the data to a spreadsheet?

Thanks in advance for any advice you can offer.

Sincerely,
Sam

[cxallan]

You certainly could Sam. OMERO has support for ROIs in its core data model. I'd envision a workflow that looks something like:

• Import images into OMERO
• Create ROIs in OMERO based on existing analysis tools
• Mark ROIs
• Visualize ROIs in OMERO.insight
• Store measurements from the ROIs in OMERO.tables
• Perform re-analysis by requesting regions of the image under the ROI
• Export data from OMERO.tables to CSV

I will stress that OMERO provides the framework for you to do all of this work by associating measurements, ROIs and metadata to images, by linking pieces of information together and by providing universal raw data access. Many of the implementation specific details of your analysis workflow will remain the same, or similar.

Perhaps you could give us some specifics so that we can respond more concretely? What language are you writing your analysis in? What is the source of your image data? What do you consider a large set of images? What types of analytics are you doing? Machine learning? Segmentation? Are you calculating some sort of feature vector?

-Chris

[SamWallace]

Thanks for your prompt reply.

I apologise for the lack of detail in my initial query.

By large I mean I have about 2,500 images.I realise now that for other HCS this could be a relatively small number. I can process them manually with macros in ImageJ/FIJI for one given workflow and it takes a couple of weeks from experience. The macros I've written automate most of the process, the only manual bit currently is the determination of a ROI around the cells (I need to visually check it's not a bad result) and the copying of the results into an excel table.

I have a basic understanding of programming, but no mastery of any given language. Though from what I've been reading I could teach myself python.

My image data is from a Leica SP2 microscope, two channel images. Some of them are 3D image stacks, but I will treat them using a different workflow. Using the macro language I worked a crude way of pulling out the metadata I was interested in. Most of the workflow I use comes from plugins or functions available within imageJ/FIJI, I can understand the source code of most of the plugins I use but I've not needed to write any code myself.

What I'm doing is basically a pixel counting method for determination of a given protein (channel 2) in the whole cell (channel 1). I then see how that measurement changes as a function of treatment groups (age of cells, source of the cells, etc). The reason I want to be able to test different workflows is to compare different methods of segmentation and see if the trends I'm seeing between treatment groups remains, or is a function of the workflow I've derived.

I've had a bit of a read around the documentation and I can see myself being able to replicate (after a lot of work) the workflow for a given image/cell. But I don't know how I'd then set it up to process all my images.

Thanks for your help with this.

Sincerely,
Sam

[wmoore]

What sort of processing are your ImageJ macros doing?

If you are prepared to manually add ROIs to OMERO, then the export of the pixel values from within them can be exported to a csv (excell) file is pretty straight-forward.

Here is an example 'FRAP' analysis Python script that basically does this.
http://trac.openmicroscopy.org.uk/ome/b ... b9c66b83d0
As part of the OMERO scripting service.
See http://cvs.openmicroscopy.org.uk/snapsh ... pting1.mov
http://trac.openmicroscopy.org.uk/ome/w ... rviceGuide

If you use segmentation or other processing steps, how do you currently save the output from these steps?

Will