Finetuning on dataset other than totalsegmentator

[rajanish4]

Hi,

Thanks for these models, I was wondering if you also have instructions for fine-tuning your models on other datasets (in-house or public). So far I only found instruction for totalsegmentator dataset. Please let me know if there is a way to do this on other datasets.

Thanks

[MrGiovanni]

Hi @rajanish4

We've provided three examples of fine-tuning on other datasets at https://github.com/MrGiovanni/SuPreM/tree/main/target_applications

Please let us know if you have specific questions when fine-tuning our pre-trained models.

Best, Zongwei

[rajanish4]

Hi @MrGiovanni, thanks, I have seen these, but my question is, what if I have a dataset that is not one of these, so do you have instructions for that? For instance, how to arrange them in folder and naming conventions and so on?

[MrGiovanni]

We arrange datasets as

Name of Your Dataset
    ├── 1st CT ID
    │   ├── ct.nii.gz
    │   └── segmentations
    │       ├── aorta.nii.gz
    │       ├── gall_bladder.nii.gz
    │       ├── kidney_left.nii.gz
    │       ├── kidney_right.nii.gz
    │       ├── liver.nii.gz
    │       ├── pancreas.nii.gz
    │       ├── postcava.nii.gz
    │       ├── spleen.nii.gz
    │       ├── stomach.nii.gz
    │       └── ...
    ├── 2nd CT ID
    │   ├── ct.nii.gz
    │   └── segmentations
    │       ├── aorta.nii.gz
    │       ├── gall_bladder.nii.gz
    │       ├── kidney_left.nii.gz
    │       ├── kidney_right.nii.gz
    │       ├── liver.nii.gz
    │       ├── pancreas.nii.gz
    │       ├── postcava.nii.gz
    │       ├── spleen.nii.gz
    │       ├── stomach.nii.gz
    │       └── ...
    ├── 3rd CT ID
    │   ├── ct.nii.gz
    │   └── segmentations
    │       ├── aorta.nii.gz
    │       ├── gall_bladder.nii.gz
    │       ├── kidney_left.nii.gz
    │       ├── kidney_right.nii.gz
    │       ├── liver.nii.gz
    │       ├── pancreas.nii.gz
    │       ├── postcava.nii.gz
    │       ├── spleen.nii.gz
    │       ├── stomach.nii.gz
    │       └── ...
    ...

You should be able to run the fine-tuning code with minor modifications to the dataloader if the datasets are organized in this format.

Best, Zongwei