Help Analyzing Image for Specific Shapes & Colors (Summer Research Internship)

I'm a rising senior in college and am doing research this summer at a cancer research institution and have been asked to use Matlab. I have used it in previous, unrelated research before, and am stumped by the following task:

Time lapse pictures are taken of a cell culture in a petri dish to understand rate of cell death. Cells and the dish itself are in black and white, but when a cell dies, it becomes a neon green. I have been asked to find a way for matlab to: - Count how many cells there are in a given picture - Count how many colored blobs there are (in the picture below, there'd be 9 purple blobs)

I cannot share a picture because it is confidential to the institution I work at, but imagine it looks something like this :

Can this be done by matlab? Would you do even try to attempt this? I was given this task late today so I haven't had a chance to create any code except for a scrap of an attempt.I ideally want to load a picture, go pixel by pixel and analyze it for its RGB value. Those with a high enough 'G'value (stains are green in my experiment, not purple) will be saved in a matrix, so then I can plot them again. But I have no idea how to get teach it the difference between the cells and a background virtually indistinguishable, let alone asking how many cells there are in a given picture.

Sorry as this is probably not the type of question you get on here. Any advice would be so greatly appreciated :)

Please let me know what further information I can provide!

Answer 1

[Image Analyst]

Actually color segmentation questions are one of the most common questions we get here. That's why I've uploaded several color segmentation demos to my File Exchange: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

In addition, you can try the Color Thresholder on the APPS tab of the MATLAB tool ribbon.

In addition, you'll need to read this http://blogs.mathworks.com/steve/2013/11/19/watershed-transform-question-from-tech-support/ to figure out how to split apart touching cells.

Also see my image segmentation tutorial in my File Exchange for info on how to make measurements of the regions once you've segmented them.

[Noah Steinberg]

Thank you for your reply.

I've spent the majority of today learning the techniques of image manipulation. I have 1 side question and 1 follow up question i was wondering if you could help me with:

1) I've found your demos either a bit too advanced for me or fairly versatile. Even "image segmentation tutorial", which you state is perfect for the beginner confuses me. Out of curiosity, what do you intend for someone like me to get out of your color segmentation demos? The effectiveness declines if I use my own picture instead of one of your demos. Should I just be experimenting with useful parts of your code that I feel would get me closer to my end goal?

2) What did you have in mind with suggesting to watershed to split apart touching cells? Would this allow matlab to count the number of blobs in a picture? For some reason, I thought this process would also include removing the cells from the black background (which I don't know how that is done, but i assume it's doable). Eventually, I'd like to create a colormap progression of these pictures taken in a time lapse. I'm just not sure I see how this watershedding technique brings me to that goal.

[Image Analyst]

1) Sorry - I have lots of comments in there, but not enough for you I guess. Perhaps if you just study it some more.

2) It looks like some of the cells may touch or overlap. That's why I thought they may need to be separated into separate blobs. If you don't have a single blob that should really be two or more blobs, then don't worry about it.