Neuron counting

Count the number of fluorescent cells (tagged with GFP) on noisy background fluorescence

Description

% Matlab GUI for neuron counting
% frequency filtering to remove background fluorescence
% Otsu's method for thresholding, with user interaction
% Original image courtesy of Lan Guo

• Required Products: Image Processing Toolbox
• MATLAB release: MATLAB 7.7 (R2008b)

[Rob Campbell]

Useful if cells are well separated but won't work for images were many cell bodies are touching. This problem is, naturally, difficult to solve.

GUI not particularly user friendly since it isn't obvious that buttons need to be pressed in a particular order. I can't see how filtering parameters can be changed from within the GUI.

[TIAN ZHOU]

hi, Rob
thank you for the comments. that is what i am looking for.

i just uploaded a new GUI. hope you'll find it more user friendly.
to answer your question about the filtering parameter, it (i.e., the slide bar) changes the threshold values in the conversion from gray to BW.
if the bodies are touching, maybe the following link is useful to you
http://www.mathworks.com/products/demos/image/watershed/ipexwatershed.html

thank you

[Rob Campbell]

Thanks, I had looked at that link in the past. My problem is rather difficult because the cell bodies as small so I have few pixels to work with. Although they are touching, the pears are fairly easy to isolate because each pear is composed of a large number of pixels.

Your new GUI isn't working for me. It doesn't check whether the input is of the right type. e.g. my image was an int8 and under those circumstances the threshold slider doesn't work. I edited the code to convert the image to a double after loading but then something else broke.

You will find other users of this site consider it bad form for an author to rate their own submissions.

[TIAN ZHOU]

hi, Rob
sorry for rating my own code. it is really not my purpose. i thought you have to click 'submit' for the rating to come out. and i haven't found a way to delete it.
you are right about the data type. it is really something i should consider.
the property of the slide bar can be changed in the GUI. select the slide bar, right click and select property inspector, and change the min and max value to expend the range. you might get an error when the threshold value is negative or more than one. i am not sure if it will help.
i'll also comment out the top hat part. i have a newer version GUI but don't want to upload it since it is really a minor change.

have you considered resizing your image before processing? enlarge them so each cell will have more pixel?
 

[TIAN ZHOU]

have you tried distance transformation before WST? there is a Matlab function BWout=bwdist(BWin).
i got this from my class notes by Prof. Alex Hartov