How to create ground truth images?

I am working on medical images we have our own medical image data. While I want to compare different de-noising filters or segmentation algorithm I needed some ground truth images. Is there exist such techniques that can generate ground truth images. If so please share link.

Answer 1

[Walter Roberson]

If you do not have data through some other mechanism (e.g., perhaps you had an x-ray as well as MRI), then you basically get someone who is skilled to segment by hand, and use that as a reference.

[Jhilam Mukherjee]

Thanks Walter my problem of segmentation algorithm is solved. But when I want to evaluate the performance of denoising filters it can't be possible manually. I need a technique that can create a noise free images for CT images.

[Walter Roberson]

Phantoms. That is, you need to create (by computing) a noise free ideal image to compare against, which you would then add noise to and run the noisy versions through your denoising algorithms to evaluate the success of the algorithms.

[Image Analyst]

Wouldn't that be nice? Unless you can use some sort of artificial phantom, you'll have to just deal with what you've got. There are denoising methods you can apply that will reduce variation, but unless you know ground truth, how do you know that what it did was correct? Can you use a phantom? If not, you'll just have to go with an expert opinion like Walter said.

[Jhilam Mukherjee]

Thanks Walter and Image Analyst. I am not using phantoms previously. I obviously use phantoms for this purpose.