Thresholding CT scan-images using imageJ/ boneJ

[dimi]

Hey! I am a 3rd year intercalating medical student and I am doing my
project on cortical development in fetuses. I am quite new with using
imagej/bonej and I was wondering if there is any function i can use to
find the threshold value for my scans. if there is no any function can
you please suggest me what is the best method for thresholding my
images? I am currenlty comparing the half maximum height (HMH) method
with the global Threshold method ( I hope I make sense) to see what
would be the most accurate method that allows me to separate cortical
bone from trabecular bone and air..

Any help will be greatly appreciated

Regards
Dimi

[Michael Doube]

Hi Dimi

 

find the threshold value for my scans. if there is no any function can
you please suggest me what is the best method for thresholding my
images? I am currenlty comparing the half maximum height (HMH) method
with the global Threshold method ( I hope I make sense)

There are quite a few automatic thresholding algorithms in ImageJ, in the drop down list labelled 'default'. They are described in the manual:
http://rsbweb.nih.gov/ij/docs/guide/userguide-25.html#sub:Threshold...[T]
http://fiji.sc/Auto_Threshold

You should probably use the Stack histogram option - this calculates the threshold based on a histogram made from all pixels in the stack and not just the current slice. BoneJ implemented stack histogram early on, but now it's in the core ImageJ so you don't need to do anything special.

The algorithm you use is a matter of judgement. Does it segment your features of interest as well as you would do it manually?

Michael

[dimi]

Thank you for your reply.. We seem to get more representative images
using the half maximum height (HMH) method than using the stack
histrogram so my supervisor and I decided to carry on with the HMH..
again thanks for your reply :)

[Abel, Richard L]

Dear Michael,
The scans are low resolution (100 microns) and there are some artefacts. Consequently the stack histograms (i.e. frequency distribution plots) do not have very distinct peaks. We will compare the HMH and Histogram values and see how similar they are.

Best Wishes
Richie

_____________________________
Dr. Richard L. Abel
Lecturer in Musculoskeletal Science
Imperial College, London.
________________________________________
From: bonej-users-a...@googlegroups.com [bonej-users-a...@googlegroups.com] on behalf of dimi [demetri...@msn.com]
Sent: 11 March 2012 13:06
To: BoneJ Users and Developers
Subject: [BoneJ: 132] Re: Thresholding CT scan-images using imageJ/ boneJ