Data

[EliasVansteenkiste]

Where can we find the data?
There seems to be a lengthy process involved to get access?

[libfun]

You can apply for the access to data here http://adni.loni.usc.edu/data-samples/access-data/
They tend to give access in about 3-4 days after the application.

[EliasVansteenkiste]

Thanks! I applied to get data access.
Hopefully they reply soon. How long does it take to train the models?

Do you perchance have any of the trained models available?
I am planning to apply transfer learning to solve a different medical problem.

[libfun]

Sorry, we don't have trained model weights, but we intend to get these by ISBI presentation of our paper. We'll make sure to share them here and notify you.

The vgg-like model takes few hours to train on five folds of cross-validation for one binary classification problem. VoxResNet-based model takes 3+ hours to train for one fold of one binary classification problem.

[CP121]

I would also be very interested in obtaining weights for your trained models. Do you have a rough ETA for this?

In your paper, you also mentioned the potential for using less preprocessed data. I am particularly interested in MRI scans that are not spatially normalised. Do you have any further thoughts or preliminary data on this?

Thanks very much, very interesting study.

[homingZhang]

Dear Friends,
My questions are as follows:
Firstly, I wanna rerun the classification results u did in the paper, so could u share with me about your ADNI data you used in the research?
Secondly, What preprocessing step did u do for the AD data ? (I saw the Input size of image was 110X110X110, so did u do some preprecessing step)
because the final results are very relevant with the Inputting data. So, I really hope u could give me some advice.