Trabecular Spacing/Thickness Issue

Hi y’all,

I’m trying to measure trabecular spaces in my bone. As previously suggested, I’m filling the background to limit the negative, background space to the interior of the bone to cut down on processing time.
(Process -> Subtract background (25 pixel rolling ball, disable smoothing). Binary -> Make binary (default binarization method and default background color).

However, after subtracting the background and binarizing the image, the thickness and spacing functions appear to be flipped, at least in the thickness/spacing maps that are generated. For example, if I tell BoneJ to calculate thickness, it’s appears as if it’s calculating spacing (thickness of the black space) and vice versa if I ask BoneJ to calculate spacing.

This reversal of thickness/spacing doesn’t happen with images that have only been thresholded and not backgrounded subtracted.

Pictures are included to help illustrate my point.

So, my questions: is there a reason this is occurring and what can I do to fix it? Alternatively, since my thickness maps are indicating that trabecular thickness is measuring the spaces in my image, can I use this value to report spacing? I just want to be sure the numbers BoneJ is giving me are actually measuring what I think it’s measuring.



This algorithm assumes that the foreground (255) of the input image represents bone and the background (0) represents spaces. It’s possible that during your processing the LUT has been inverted, so that the foreground is black. The actual maths used for Tb.Th and Tb.Sp is identical, just the pixels used are in the complementary foreground/background phases.

Thanks for your reply, Michael.

The binarized images I’m producing have an outline around the edge of the bone, which as far as I can tell is due to the rolling ball threshold value.

This noise around the edge of the bone makes a difference in my spacing mean and SD. I’m currently using the despeckle function to get rid of the noise.

Is the despeckle function the best way to resolve the noise on the rim of the bone? Or is there a better way to do it?

Try some other filtering and segmentation approaches until you get images that properly represent the parts of the specimen you want to measure. Median filtering can help to clean up boundaries, prior to thresholding.