Hi all,

I am working on a multiple organ localization problem in medical image volumes using random regression forests (RRF). Our work is based on the method proposed by Criminsi et al.

Let us take the scenario where we try to localize the right and the left kidneys. Let us also assume that we have a good trained RRF that does a good job at localizing the kidneys.

Now imaging a scenario where we test an image where one of the kidneys of the patient is removed. And we know that RRF gives us good land marks (hip bones, spinal column, etc.) to locate our organs of interest. Since the landmarks are still there, the testing voxels will most likely get split in the same manner as in the case where both kidneys are present. This will lead to localization of a non-existing kidney.

I am wondering how can I detect the existence in such a case before doing the localization?

I am doubtful whether I will be able to do a good job at detection in similar scenarios unless organ specific details are incorporated in the training phase.

Thank you in advance for your inputs.

http://research.microsoft.com/apps/pubs/default.aspx?id=179759

More Prasad Samarakoon's questions See All
Similar questions and discussions