@djpbarry, I’m working on learning ADAPT but initially having some problems with segmentation SNAFUs. I think I can resolve this experimentally by lowering my cell density but wondering if you have any comments on troubleshooting segmentation. https://imgur.com/a/NEcuO48
“The cyoplasmic signal should consist of a bright cell (or cells) against a dark background - phase contrast images, for example, will result in poor segmentation results.”
This is most likely why you are running into issues… Now, I’m not sure if you are able to:
- re-take your images using fluorescence labeling? Are you married to bright-field acquisitions?
- do some pre-processing on your images to result in a binarized mask to use as input in ADAPT. But with that - I’m not sure you can even do using this plugin… might be worth a try? @djpbarry ?
Though too - your images are a bit out-of-focus and over-exposed… might be worth getting back on the scope and doing a few new acquisitions while teasing out an analysis pipeline.
- You can engineer your cells to express a fluorescence marker and reacquire images.
- Or alternatively, you can pre-process the images you’ve got, either through filtering in #fiji or perhaps doing some pixel classification with #ilastik, to generate “psuedo-fluorescence” images to use as inputs.
Thanks @etadobson @djpbarry. Eventually I want to colocalize morphology data from transmitted light images with endogenous NADH fluorescence, so working with the GFP version of these cells would interfere with that analysis, I think. It’s true that I need to pre-process. I will play with ilastik.