Do you guys have any plan to support DICOMs?

This tool seems very promising and it’s currently very hard to find any tool to use for Medical Imaging. In this area, we usually work with DICOMs, which has the image itself and some tags. These images usually are 12 or 16bits and the radiologist has to do some windowing to convert them to 8bits. There are a lot of DICOM viewers, but no open source solution to annotate DICOM images.

1 Like

Hi @fernandocamargoti, glad you find it interesting and thanks for the prompt!

I will have a look at how we could integrate DICOM, and get back to you.

Let me ask you - what other information besides the images is important to you? e.g. are the tags important?

There are some information on the tags regarding the patient that might be useful for the radiologist (like sex and age) and possibly other tags might be important for him to see too (like the patient’s position). So, if there was a way for them to see the DICOMs’ tags, it would be cool too. But the main thing is really to have a way to change the windowing of the image. Have a look here, for an example:

https://www.kaggle.com/dcstang/see-like-a-radiologist-with-systematic-windowing

This notebook shows the huge difference in appearance of the image with different windows and shows that, during the analysis, the radiologist uses multiple windows to see different things.

About implementations, I found these JS libraries:


And to manipule DICOM within Python, there’s the excelent PyDICOM:

Thanks a lot for the links, I had a look and tried few demos. Some questions to help me understand better:

  • I see there are 2 dimensions across which windows can change ( brightness/contrast, or window center /window width), is this correct and standard across DICOMs?

  • Would you want to have the annotation to be persistent across windows or you need option to reset it?

  • For exporting the annotations once drawn, do you have any special preferences? e.g. file format, information to export (such as window level annotation was drawn)

  • It’s common to use window center and window width. They naturally end up changing the brightness and contrast. Like, if you have a large window width, you have less contrast, if your window center is low, you have more brightness, and so on.
  • The annotations are unique for the image. The windowing is just for the radiologist to be able to analyze the image itself. In theory, we could export the image to PNG with a given window and let the radiologist analyze this image. For X-rays, it would be ok, but for CTs, MRIs, and other images, the radiologist needs to see the image in a lot of different windows. But the final annotation is the same.
  • I think the same way the annotations are exported right now.
1 Like

Great, thanks a lot.

So taking this as an example:

Image 1 to 151 would appear in Remo as different images. Each image would have its own annotations, and we can add some scrollbars or mouse shortcuts to easily change windows.

Do you think this could be a working solution?

If yes, it would be quite easy to add and we could release it in the near future.
Adding the export for PNGs would be extra work, which would add at a later stage

Yeah, correct!

The only thing that I’d add is a way to group these images. I mean, the radiologist usually looks at all the images of a given series (for example, CT might generate 500 images). If the images could be shown in order, that would be great.

About the PNG exportation, it’s not necessary. I was only giving an example when I talked about PNG.

1 Like

Ok great, thanks.

In the first iteration, the way to group images will be rudimental:

  • sort them by filename
  • or add different DICOM to different datasets

Later we are planning to add folders, which will naturally allow to group images within a dataset.

We are establishing the priority of the ticket, will let you know once it’s out.

Sure, that sounds great for me.