search

uni-medical / STU-Net

The largest pre-trained medical image segmentation model (1.4B parameters) based on the largest public dataset (>100k annotations), up until April 2023.
Apache License 2.0
264 stars 24 forks source link

Fine-tuning models with nnUNet v2 using multimodal data #31

Open SergioRodLla opened 1 month ago

SergioRodLla commented 1 month ago

Hello,

Thank you very much for the repository and the provided pretrained models. I have seen that you added some scripts for nnUNet v2 to run fine-tuning as well as the STUNetTrainer. I have a dataset with CT and PET modalities and I was wondering if it is possible to use nnUNet v2 to run fine-tuning using a pretrained model that you provide (e.g. huge_ep4k.model) to do this combining both modalities (concatenating them as input as you showed for the AutoPET challenge).

Best regards, Sergio

Ziyan-Huang commented 1 month ago

Hi Sergio,

Fine-tuning on multi-modal datasets like AutoPET (CT and PET concatenated) is supported. We copy the first convolutional layers (pretrained on a single CT modality) from the pretrained model to support multi-modal inputs.

Best regards, Ziyan