Question- Capability of analysing a large number of X-rays for specific areas/nodules of decreased opacity

Hello,

Sorry for the bother, especially if the answer to this question is already readily available. Next year i have plans to begin a project which will involve the analysis of a large group of Chest X-rays.

This will involve looking at a series of X-rays and looking for the presence of areas of decreased opacity, specifically looking for areas of pulmonary interstitial emphysema. This research will require me to analyse a large number of X-rays and to count the number of these areas present in the images, as well as the diameter of the nodules.

I am wondering if the ImageJ software and various available plug-ins are capable of automatically analysing the series of X-rays, and to automatically analyse the number of black nodules present as well as their dimensions, or if i will be required to manually analyse these image myself.
And if so, how this can be accomplished- what plugins are required, and what tutorials or guides to perform these actions are available.

I hope this question makes sense and sorry if its already been answered.

Thank you very much.

Good day!

No problem, because it is very rare that an identical request is posted but of course there are apparently similar ones that most often turn out to be answered in different ways …

So here we go:

Without representative images we can’t help you at all!

  1. Please supply original, i.e. un-processed, X-rays as 16bit TIF-images. Because you won’t be able post those directly to the Forum, make them accessible accessible via a dropbox-like service.
    (X-ray image showing 8bit depth are generally unsuited!)
  2. Explain exactly what is of interest for you and what, if possible, should be detected or extracted or classified autoamtically. Yoiu may use a copy of one of the images where you outline or otherwise identify by drawing on it “what is what”.

Regards

Herbie

Thank you very much for the reply.
Images attached are images of similar X-rays to what i will be looking at next year. They show pulmonary interstitial emphysema- as part of this condition, there will be areas of the x-ray in which there is increased lucency/increased blackness where lung tissue should be.
The image shows pulmonary interstitial emphysema (PIE)- the arrows on the second image area pointing towards areas in which the PIE is present, whereby there is an area of increased blackness. These areas are rounded or linear in shape. The main thing that i would like to be able to extract would be

  • The number of these areas present within the lung
  • Specific parameters of these areas, such as their diameter.

The proposed project would involve the analysis of a large number of X-rays, potentially 200 or more in which there are more significant signs of PIE on the images, and as many as 800 where there may be no signs of it, or a small number of these areas present.

These images are from Radiopaedia, a useful learning resource for radiology. If these images dont appear in this response, they can be found from this page
https://radiopaedia.org/articles/pulmonary-interstitial-emphysema.

Once again, i hope this makes sense and sorry for the bother. I have only just begun to look at ImageJ and am still not entirely clear on its capabilities or how to operate it properly. I really appreciate the help.

Thank you very much for the help.
The image provided was unfortunately JPG, however in the actual research raw images will likely be used. This image was also probably not the best example, however i currently do not have access to the real X-rays on the hospital system at this point. However, fortunately with those images it will be possible to alter the intensity of the images and help make borders more clear.

I have been trying to familiarise myself with the guidebook for image J and Fiji, but am still having a bit of trouble. The image example that you provided in your reply looked like a good example of what i’ve been trying to do, despite the limitations of the image i provided you.
I was just wondering if you would be able to give an overview of what commands and changes you made to the image to get to that point, in particular what threshold changes and analyse particle settings you used. I would love to try and replicate this and a step approach would really help give me an example of how i can use the software as desired.

Thank you very much for all your help, i really appreciate it.

Good day!

I have been trying to familiarise myself with the guidebook for image J and Fiji, but am still having a bit of trouble.

Please be more specific about your “bits of trouble”!

I was just wondering if you would be able to give an overview of what commands and changes you made to the image to get to that point, in particular what threshold changes and analyse particle settings you used.

It is far from reasonable to tell you the processing details, as long as you don’t provide the actual 16bit images. The processing depends strongly on the image contents and quality.

My present approach is based on an individual analysis of the original image histogram:
Histogram%20of%20pulmonary-interstitial-emphysema_excerpt
The original histogram is automatically windowed which results in the following:
Histogram-2
To the corresponding image a band-pass filter is applied, matching the expected size range. Finally the resulting image is thresholded according to the automatic “Huang”-scheme.

I don’t think that my present approach generalizes to the actual 16bit images that are urgently required to give you relevant advice.

Regards

Herbie

you could look at trying the hough circle “finder” (sorry not sure on exact name or url) but you can google it.

It should with some tweaking be able to match the nodules for you, if they are relatively circular (should be in most cases).

…just one idea

if they are relatively circular (should be in most cases).

From what I’ve analyzed using the provided preliminary data, they are not.

Regards

Herbie