Request for AAPM CT checkpoint in the paper.

hyn2028 / tpdm

Official code for "Improving 3D Imaging with Pre-Trained Perpendicular 2D Diffusion Models" (TPDM)

47 stars 4 forks source link

Request for AAPM CT checkpoint in the paper. #6

Closed ECHOEeyes closed 5 months ago

ECHOEeyes commented 7 months ago

Hello, your paper showcases excellent reconstruction performance of sparse view CT, which has piqued my interest. But what is provided in the code is a checkpoint for another set of simulated data, not for the AAPM CT dataset. I'm not sure if it's convenient to provide the pretrained checkpoint for AAPM CT. I also noticed that your previous work on DiffuseMBIR provided checkpoints for AAPM, and if TPDM could be provided, it would be more helpful in forming a comparison. Thank you.

hyn2028 commented 7 months ago

We apologize for the long wait. Checkpoints and sample volumes related to LDCT-CUBE which was used in our paper have just been released, please see the updated README.

However, The LDCT-CUBE checkpoint originated from an experiment to observe the behavior of TPDM when the dataset was extremely small (10 volumes) to train a 3D model. Therefore, its performance may be insufficient for use in abdominal CT-related tasks.