xmed-lab/genericssl

NeurIPS 2023: Towards Generic Semi-Supervised Framework for Volumetric Medical Image Segmentation

genericssl's Issues

nothing

nothing

recurrent

Hello, I conducted an experiment on the m&ms data set according to the DiffUNet.txt you provided. I can achieve an accuracy of about 72% on 2% of the D domain, and an accuracy of about 69% on the 2% of the A domain. , the effect is very different from what you said. I would like to ask if you set some specific parameters when using unet as backbone training?

train_diffusion_2d.py

Hi, thanks for the code. I would like to ask, do train_diffusion_2d.py and train_diffusion.py in your code handle 2d slice data and 3d data respectively?
Is there a big difference in accuracy between running train_diffusion_2d.py and train_diffusion.py?

3d scgm

Hello, I would like to ask if you have a way to process the scgm data set that your network can receive? The scgm data sets I processed before were all processed into 2D slices. Can I directly modify your preprocess_mnms.py code? Looking forward to your reply, thank you

Request more detailed experimental parameters

Hi,

Thank you very much for your excellent work. I would like to ask for a guide on using the dataset mentioned in your paper. I have attempted multiple rounds of training on the Synapse dataset and tried different epochs, but found that the accuracy of the reproduced results still differs from the data in your paper. May I ask if it is possible to provide more detailed parameter explanations so that I can better reproduce your results?

Thank you very much for your help!

Running with other dataset

Hi,

Thank you for your great work. Would it be possible to provide a quick guide on how to run this code with different datasets other than the four mentioned in the paper?

Thanks!

Request more detailed experimental parameters

Hi,
Thank you very much for your remarkable work. I've conducted experiments on the LA dataset using the configuration consistent with Table 8 in your paper (same with the config file). The obtained results show some variance from those reported in the paper. Could you please advise if there are any additional parameters I should adjust to better replicate your reported results on the LA dataset?

Thank you very much for your assistance.

backbone

Hi, thank you for sharing the code. Does your train_diffusion_2d.py use unet as the backbone?

SCGM dataset

Hello, I would like to ask if you have done experiments on the SCGM dataset?If so, can you share your experimental results on this?

The results of training dice on the new data set

Hello, sorry to bother you . I have processed the scgm data set and trained it on the genericsl code you provided. This is the result displayed during the training process:


But the dice value of the running result is very low. Is it because there are too few samples in the scgm data set? This is using 20% of the scgm dataset train_toA_labeled_0.2.txt file, which has only 6 site2-sc03-image site2-sc07-image site3-sc03-image site3-sc08-image site4-sc04-image site4-sc07-image.If you have time, I hope you can give me some suggestions and I look forward to your reply. Thank you so much!