search

mobaidoctor / med-ddpm

GNU General Public License v3.0
153 stars 16 forks source link

Using Whole-head MRI synthesis pretrained model for 2D reconstruction #38

Open Sarah-2021-scu opened 2 weeks ago

Sarah-2021-scu commented 2 weeks ago

Hello,

Thank you for your work. I am trying to use your pretrained Whole-head MRI synthesis diffusion model for 2D image reconstruction. My image size is (256, 256). Apart from resizing my image to (128, 128), what changes would you suggest me for the fine-tuning process?

Thank you in advance.

mobaidoctor commented 2 weeks ago

Hi @Sarah-2021-scu, thank you for your interest in our work and for reaching out! If you're working with a 2D dataset, we highly recommend using the 2D version of Med-DDPM, called Polyp-DDPM (https://github.com/mobaidoctor/polyp-ddpm), which we specifically designed for 2D medical images. Med-DDPM is tailored for 3D images, so its architecture isn't suitable for 2D images. Polyp-DDPM, however, is optimized for 2D images, resulting in faster convergence due to the lower dimensionality compared to 3D models. This 2D version provides diverse, high-quality results and converges quickly, making it an ideal choice for 2D images. If you have further questions or need any clarification, please feel free to let us know. Best, Zolnamar