search

Yaziwel / Region-Attention-Transformer-for-Medical-Image-Restoration

[MICCAI 2024] Region Attention Transformer for Medical Image Restoration.
Apache License 2.0
22 stars 2 forks source link

about data preprocessing #2

Open CyberHuman777 opened 2 months ago

CyberHuman777 commented 2 months ago

Hello! Because the value range of PET Image, CT and Pathological Image are quite diffierent. Have you do any preprocessing? And how to do it?

Yaziwel commented 2 months ago

Hello! Because the value range of PET Image, CT and Pathological Image are quite diffierent. Have you do any preprocessing? And how to do it?

To accommodate input for the SAM model, I initially converted all images to 8-bit for both SAM input and image restoration in the experiment of the paper. However, I now realize this approach may not be ideal. We only need to preprocess the images to 8-bit when inputting them to the SAM model for obtaining region partitioning masks. The images should then be processed within their original data range.