Moments of Inertia / Slice Geometry

Hello everyone,

I’m currently working on a clinical trial where I need to measure shadow plugs on a femur from an MRI. This is why I wanted to use Slice Geometry. As it is a clinical trial, I cannot show you the images.

While I was trying to understand BoneJ’s functions, I saw that it was better to use Moments of Inertia before to align the bone to the principal axes.

I tried to use Moments of Inertia but the results were really awkward. It was really not what I expected and maybe it is because I’m new to BoneJ :woman_shrugging:

So, I started over and drew a ROI on the shadow plug, then I directly used Slice Geometry and got some results.

Do you think I can use those results even if I didn’t used Moments of Inertia ? As I am only using Slice Geometry on the shadow plug (where there is the ROI) I don’t know if Moments of Inertia is really useful and if I can get rid of it as I have difficulties to use it.

Thank you in advance for your answers !!

Please check the regulatory requirements for software used to analyse your clinical trial data. We have not sought regulatory approval and BoneJ has no warranty in any jurisdiction, in common with practically all ImageJ plugins including ImageJ itself.

If it is OK to use BoneJ for your clinical trial, then:

What sequence are you using? Typically bone tissue has low signal in MR images due to having tightly bound protons, so it usually has a low pixel value and looks dark on the images. Bone marrow typically has a high signal, as does surrounding muscle, vessels and fat. BoneJ was written and used with X-ray attenuation contrast in mind (CT, XMT, SR-µCT) in which bone has a high signal relative to other tissues and air.

MR images usually have highly anisotropic pixel spacing, with much higher resolution (smaller pixel spacing) in-plane than between planes. Your advantage is that human patients are typically positioned with the limbs parallel to the gantry axis, so the femurs should be nearly aligned already.

Given the limitations above, I’d recommend in your case not to use Moments of Inertia. You’re better off working on untransformed pixels and taking really simple measurements, such as pixel count (≃ volume), point-to-point lengths, and so on.

If you need to use Slice Geometry, you’ll have to specify the pixel range that relates to bone. In your MRI case it’s likely to be a low pixel value, whereas the defaults in BoneJ assume a high pixel value typical of X-ray attenuation. You’re also likely to have to segment bone from other low-signal regions (tendon & ligament, air). Please pay close attention to these issues otherwise your measurements may be meaningless at best and misleading at worst.

Can you give more detail on the kinds of measurements you need to take? Please spend some time really getting to know your images and what the measurements mean and how they are made. I’m concerned that real diagnostics and therapies may be developed on the back of poor understanding of the data and its analysis.

Thank you for your answer @mdoube

I should have said this first but I’m actually a second year student in a Bioengineering school and I’m working on a clinical trial underneath the supervision of my tutor for an internship. It is her who decided to use BoneJ but I told her what you said, and she said it was ok.

Thank you for your answer on Moments of Inertia it really helped me to understand ! Regarding the type of sequence that I’m using, it would be better if I say that in private and not posted on a forum, if it is okay ?

Even if simple measurements seems interesting, my tutor wants me to use Slice of Geometry to do some measurements and compare them with data they already have. As you saw, I’m a newbie to BoneJ that is why I’m looking for help.

Can I ask you 2 questions ?

  1. I now completely understand why I need to specify the pixel range that relates to bone. But do you know how can I change the pixel value ?

  2. Do you know if there are any tutorial or if you know how to segment bone for other low-signal régions?

All the best.

From the user guide:

  • Bone Min : Lower threshold, in either uncalibrated greys or HU
  • Bone Max : Upper threshold, in either uncalibrated greys or HU

Segmentation is a large problem with many potential solutions that you need to test and customise for your particular images and research questions. It can have a very large effect on your final result, so you have to do a sensitivity analysis to check how reasonable variations in your processing steps affect the readout values.

You could start with setting a threshold (Image > Adjust > Threshold) to find the pixel value range occupied by bone. Note the other regions that have the same pixel values, and also that there isn’t a clear cutoff value (this is one simple example of what I mean by reasonable variation in processing steps). You have to exclude the other regions somehow - either with a cleverer segmentation approach or by manually deleting, or cropping. Note that Slice Geometry can be restricted to a rectangular ROI (just draw a rectangular ROI on the image prior to running Slice Geometry).

Then, let’s say that I draw a ROI on the shadow plug i want to measure (a polygon one is okay ? ). Regarding the pixel value : (Bone Min / Bone Max), how can I find the values that I should take ? Should I set a threshold as you explained or is it different as I do a ROI only on a shadow plug ?

If I draw a ROI as I explained, do you think it is still useful to do a segmentation or not ?
I don’t really understand why segmentation is that much important.

Thank you for all your answers @mdoube ! I really appreciate :smiley:

You need to check the pixel values that relate to bone. Try Image>Adjust>Threshold to get an idea interactively.

Segmentation splits the image into regions that have logical meaning, e.g. foreground/background. Please take some time to do some basic image processing and analysis reading. John Russ’ book is a good start: The Image Processing Handbook. Your university library probably has a copy of it or a similar introductory text.