search

Maartendrinhuyzen / ai4mi_project

MIT License
0 stars 0 forks source link

Find an application for affine transformation in preprocess step #4

Closed Maartendrinhuyzen closed 2 weeks ago

Maartendrinhuyzen commented 1 month ago

For patient 27, there is a ground truth image (GT) with all the highlighted parts (THOR) separated and then there is a corrected ground truth image (GT2) with all highlighted parts connected. We need to figure out how the affine transformation (rotation and translation) was done and apply it across all patients because the transformation is uniform across all of them. The TA also mentioned that the slides for Image Registration contain the way to reverse the transformation. Starting point could be elastix (slide no. 116-18)

Maartendrinhuyzen commented 3 weeks ago

maybe someone can review whether the transformation dataset is in fact correct. It was to my eye but perhaps we can apply the four eyes principle here..? @lisamenzel99 @SS-Fatima @others

Maartendrinhuyzen commented 3 weeks ago

I found my current implementation somehow copies the heart (so in some images two hearts are present). As far as I know this is not desirable. So needs to be fixed.

Maartendrinhuyzen commented 2 weeks ago

Just fixed the issue. My initial thought was correct: the previous code indeed copied the heart twice in the image. Now the previous heart is first removed before the transformed heart is pasted.