0

We are working on a problem of lesion tracking. So we have 3-d series of dicom images for two time points. Lesions are marked on time point - 1, using this we want to project the same lesions to the time point 2.

Currently we are following the approach below:

  • Register 3-d series of images from timepoint-1(moving_image) with 3-d series of images from timepoint-2(fixed_image).

  • Use the transform from the above registration to resample the masks on timepoint-1 with fixed image on timepoint-2 ==== This gives fixed_mask on time point 2.

  • Overlay this fixed_mask to fixed_image to get the lesions on the timepoint-2.

Doing the above gives all blank images as the mask is mostly filled with null except that one tiny point.

Am I following the right approach? Is there any other way to overlay the mask from one time point to the next?

Kaushik
  • 1
  • 1
  • What interpolation are you using when resampling? With a linear interpolation on a binary mask, you're mostly going to get zeros. Try nearest neighbor interpolation. – Dave Chen Nov 12 '21 at 14:39
  • @DaveChen Thanks for your response. I initially tried it with linear interpolator but then I also tried nearestneighbor after seeing your comment but it doesn't seem to work. It still returns null images. – Kaushik Nov 15 '21 at 09:37
  • can you share a sample image? – Dave Chen Nov 15 '21 at 14:26

0 Answers0