Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Why I have a low accuracy when used 3dcnn in this tutorial? #1

Open
ABadCandy opened this issue Apr 10, 2017 · 9 comments
Open

Why I have a low accuracy when used 3dcnn in this tutorial? #1

ABadCandy opened this issue Apr 10, 2017 · 9 comments

Comments

@ABadCandy
Copy link

Hi, thanks you for sharing first!
I have about four hundreds data from ADNI and follow your tutorial with three classification(AD,MCI,CN), but I have a low accuracy finally, I worder know is there a problem with my data processing problem or model? So could your please send me a sample data for study, thanks a lot!

@regnerus
Copy link
Owner

Hi!

First of all, this wasn't really meant as a tutorial but great that it was explanatory enough.

About the low accuracy: It could have to do with the dataset, the dataset I used which was uploaded to Kaggle but quickly taken offline again already was pre-processed. This page still contains the readme from the dataset which also lists the steps done in pre-processing. https://github.com/jbrown81/ADNI_kaggle/blob/master/adni_kaggle.ipynb

@ABadCandy
Copy link
Author

Thanks for the reply!
I also read the pre-processing steps and other papers before , then I use the FSL software to process the raw data and get the gray matter slices in coronal plane.
However, when I use the pre-processed data to train a 2DCNN or 3DCNN both don't have a great result.
Maybe can you send me some sample pre-processed data to [email protected] ? Thank you for helping!

@regnerus
Copy link
Owner

I have a zip file with the pre-processed images on Google Drive. https://drive.google.com/file/d/0B5zQ7kc5shlgRG9laXMxcnlyMk0/view?usp=sharing

Good luck!

@ABadCandy
Copy link
Author

Thank you very much! I will try again~

@regnerus
Copy link
Owner

@ABadCandy did it work? 😄

@ABadCandy
Copy link
Author

Yes! I used the pre-processed data you provided on 2DCNN and 3DCNN , the accuracy is ~80% and ~85% respectively. However, I also want to use the original data for comparative experiment, Do you have the raw data downloaded from adni?(I cannot download it now)Sorry to bother you!

@regnerus
Copy link
Owner

@ABadCandy no sorry I do not have that data available but you should be able to download it somewhere right? http://adni.loni.usc.edu

@ABadCandy
Copy link
Author

@regnerus Yes, but I cannot find the original MR image for each person correspondingly and it seems that the data on website has not allowed to download. Anyway thank you very much!

@czainab
Copy link

czainab commented Apr 17, 2018

Hi @regnerus i am trying to run your code but I am having an issue can you please share the weights file and report. It will be of great help for me.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants