search

StanfordMIMI / DDM2

[ICLR2023] Official repository of DDM2: Self-Supervised Diffusion MRI Denoising with Generative Diffusion Models
125 stars 20 forks source link

Training Dataset question #34

Open kklucy11 opened 2 months ago

kklucy11 commented 2 months ago

Hi, thank you for this amazing paper. I wanted to ask you very few questions to elaborate in very detail. Should the data used for training during the training process be 1 4D data? Can a model trained on only one 4D data be used on other people's 4D data?

tiangexiang commented 2 months ago

Yes. DDM^2 is a self-supervised optimization-based denoising algorithm. That means the training data needs to be the testing data itself. With this being said, we require one single 4D MRI acquisition (more volumes are better) as the input. After training, the trained model is able to (and is only able to) denoise the same 4D MRI acquisition.