Home | Mission | Network Members | CMCI Seminar | Courses | Downloads|
Weblog | FAQ | Links | Contact | Calender



EMBL-CMCI Questions & Answers    (last update 8.12.2005)



Contents

Image subtraction to measure growth of neurite
























Image subtraction to measure growth of neurite (ImageJ)
Question
I'm having problems with image subtractions and I was wondering if you
could troubleshoot me.
(You have to be patient with my ignorance on Image Processing):

This is what I've to do:
To measure the dendritic growth of plated neurons (labeled by DiOlistics)
by timelapse imaging.

This is what I've been doing:
Once i acquire my video, I videotrack each growth tip (I image several at
the same time) to measure how much it moved.
But This is time consuming as I'm only interest in a gross measure of growth.

As an alternative, I was thinking of just take the last acquired frame and
subtract it to the first one of timelapse.

So this is what I've been doing with ImageJ:
1) Acquire the Timelapse (Metamorph software)
1) Extract the 1st and the last frame of the timelapse stack (they are
16-bit images)
2) Align the two images with the ImageJ Plug-In TurboReg
3) Subtract them using ImageJ Process->ImageCalculator

But there are somethings that I don't understand:
1) Why the subtracted images have always positive intensities? Is this a
Java thing? Would MetaMorph be better?

2)what is the best way to extract information from the subtracted image?
- the number of black & white pixels?
- Intensities outside Mean+/- SD?
- A threshold image?

Any comments or hints you can give me for sure will be very useful.

Answer
it seems that you have alot of knowledge about processing so I have not much to say. My suggestion would be:

(1) Use thresholded image (black and white). Then down scale to 8bit. Dendrite white, background black.

(2) Black pixels contain intensity 0. so if you subtract white from black (0-255) then the result is 0 in 8 bit or 16 bit because their available numbers are between 0 and 255 (8bit) or  0 and 65536 (16bit), no negative numbers. *1

(3) for the reason (2), when you do image calculation, select "MIN" instead of "subtract". And also before do ing the image calculation, invert the first frame. This will leave the grown part (white, 255) only.

(4) measure the white area in the resulted image. This can be done by using "wand tool" in the menu bar to select the white region and then "measure".

I hope this helps. If you couldn't get a result, send me the first image and the last image.

*1 This is same for all software.

8.12.2005