I have this kind of microscope images, I wanna to get the average yeast cell size. But I don't wanna measure them manually. I tried use the particle analyze function in imageJ to analyze the images, but it doesn't work very well. So I am thinking about developing a simple machine learning program to recognize the cells in these images. It may take more time comparing with analyze them manually. But I can learn something new.
So I am wondering where should I start with? If you saw any program with similar purpose, can you share the link? So I can make some modification base on it then it may fulfill my purpose. Thanks.
Dear Jingjing Gong ,
Your image looks quite nice, so I don’t expect many problems in trying to segment them with a classical pipeline. You can change your image to RGB, and split colour channels in order to separate stained from non-stained cells. To segment you can probably use something along the lines of gaussian filtering -> top-hat filtering -> thresholding -> remove small and too large objects. After that I think you should get quite good numbers. See attached what I could obtain by working quickly on the green channel of your image using that pipeline.
Now if you are interested in a simple machine learning method I would suggest starting with the Weka Segmentation package that is available for ImageJ/FIJI.
https://imagej.net/Trainable_Weka_Segmentation
- Badges
- Science topic
- Similar topics
- Elementary Particle Physics
- Particle