![]() ![]() Of course, this is a very niche case! Just wanted to see if it was possible, perfectly understandable if it isn’t. So this would have to be an image-based analysis (was imagining FilterObjects by the desired class to isolate nuclei of this cell type, then IdentifySecondaryObjects or similar to then essentially do a marker-based watershed to obtain cell masks for shape measurements), rather than being based on existing measurements unfortunately. But after having used a trained classifier originally to classify nuclei across the multiple classes of cells in these cultures. The main thing I have in mind is morphology analysis on a particular cell type (for which the cells were originally stained for a membrane stain for that cell type). Yes it would be easiest to have it all in one continuous pipeline, but as I’m training a model on the dataset for classification then unfortunately that’s not possible for my case. csv somewhere, then is it possible to load the objects and classifications back into CellProfiler for follow-up analysis? Or would I need to take the original segmented objects, make all measurements again, classify again but within CellProfiler using the saved model, and then FilterObjects etc etc to perform that type of analysis? If the classification was done in CPA (or CellProfiler for that matter) and stored in a. I’ll add, though, that for a future analysis I think I would want to take these classified objects, filter them for a particular class, and then do follow-up analysis of (for example), morphology of that particular cell type. However, you’re absolutely correct that I could just run the classification directly in CellProfiler Analyst at that point, and I was overcomplicating it - due to the nature of some previous work, I am used to using CPA as a training tool for a model rather than running the analysis on the whole dataset in CPA as well. And then, after training a model, be able to feed these back into cellprofiler to use ClassifyObjects on with that model. I think my original idea is that I would have measurements made via the first pipeline for every object. If LoadData won’t work, then I’ll have to rerun all the measurement modules every time I want to iterate on the classification step and check overall progress…Īny help would be really appreciated! Thanks In addition, I would like to take a look at the different batches of this experiment that pop out prior to doing a final analysis with a shared analysis of all batches with same classifier at the end. ![]() I would like to avoid this if possible, as it will be a very large dataset and doing this will add a lot of computation time on top of the considerable amount there will be already. I’m hoping it is possible, as otherwise the only solution would be to reload the images and saved objects from the first pipeline, which should be straightforward with the normal input modules, and then rerun all of the measurement modules before applying the classifier. As I want to load measurement data, I think I need to use the LoadData module - but is it even possible to load Object data this way? The documentation says that per-image data can be loaded, and that images/object images can be loaded, but does not mention per-object measurement data. I already have all of the above steps working great, apart from loading objects and their associated data back into CellProfiler in ‘pipeline 2’. After having used those exported object measurements to produce a classifier in CellProfiler Analyst (using only a subset of that data), I would then like to load all saved objects and their associated measurements into CP and apply the classifier with ClassifyObjects.Segment cells and make object measurements (MeasureSizeShape, MeasureIntensity etc), and then save the resulting objects and measurements.I am trying to design CP pipelines to perform the following steps:
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |