Ideal threshold method for trabecular volume analysis of microCTimages

I have been using bonej to assess trabecular structure of mouse tibia. Images are obtained with Bruker microCT, reconstructed at 10microns.

I have been using BoneJ1 as I can generate a 3D reconstruction of the trabecular bone surface.

I have noticed that default ImageJ threshold is often to strict. (Not picking up thin segments of bone). Huang seems to be the best option. (Although it probably over estimates bone thickness)

I use do the thresholding across the entire stack. (Stack Histogram option). Also I try crop the image to reduce the amount of “background”.

I was wondering if there is any hard or fast rules in terms of ideal thresholding methods for trabecular bone analysis.



Short answer is no, there are no hard and fast rules. The best thing for you to do is try some different approaches and measure your outcome variable to see how it varies as a function of your thresholding or other segmentation technique. Sometimes measurements are robust to small changes in segmentation, other times they are very sensitive. It depends on what you’re interested in and the nature of your images.

BoneJ2 also does this, but you have to first save an STL then open the STL in a 3D viewer.

1 Like