ha-ha-ha-han / ukbiobank_deep_pretrain Goto Github PK
View Code? Open in Web Editor NEWPretrained neural networks for UK Biobank brain MRI images. SFCN, 3D-ResNet etc.
License: MIT License
Pretrained neural networks for UK Biobank brain MRI images. SFCN, 3D-ResNet etc.
License: MIT License
Hello, i'm having trouble replicating results with your model. I'm using IXI database, and i'm getting MAE around 8 using the example notebook u published. Do u have any idea why is this?
Dear @ha-ha-ha-han ,
Congratulations on your model and accompanying article!
I would like to use it in my research, which is also on UK Biobank data, but have had some difficulty reproducing the same level of performance in the paper. Namely, in a sample of 2381 UK Biobank T1s, I get a mean absolute error of ~5.05yrs, with a notebook based on your example.ipynb available here.
From reading the paper, I think I am using the correct data in the correct formulation (namely the T1_brain_to_MNI.nii.gz's, recorded age - Data-Field 21022, and your model as depicted in your example.ipynb file,but would be most grateful for your guidance!
Best wishes,
James
Hey, this is great. Thanks for making it available.
Can you provide an actual brain image example in addition to the random noise example?
Dear authors:
Recently I have a looking at your this project and feel interested in it ! I am interested in this project, but I also face some challenges. and I wish I could receive your suggestions.
The question 1 is that, in the site: #1 , you refer to that, "We transferred the model to the PAC2019 challenge data (multisite, different age range than UKBiobank, 16-90yrs) and it saved some training time if you use UK Biobank-pretrained weights as an initialization. " and here, I want to know, how could I get the Biobank data for using in this project? And the same, the PAC2019 challenge data?
Thank you. I am looking forward to your reply Sincerely!
Best regards.
Hi, han. I'm really interested in your method and I have met some questions.
When it goes to fine-tune task with other datasets, the loss gets lower, but the mae or mse holds.
I used the kl_loss from your code as the loss function
Hello, I am trying to replicate your SFCN brain age prediction model to do my undergraduate thesis on top of it. My idea is to use your model that has great results and retrain it with part of my data and your input data. But I'm having some problems with the code, because machine learning is kind of new to me, could you share the code that was used to train the model or part of it, some parameters, anything would help? Thanks.
Hi @ha-ha-ha-han - thanks for sharing example code and pretrained model! I am trying to replicate these findings and currently seeing promising results. I do have couple of questions which I am hoping to clarify before proceeding further.
First, the published article reports results from an ensemble of models. Does this mean there are additional pre-trained models (20?), whose predictions are subsequently averaged? If so would you share those as well?
Second, do you also have a code snippet for bias correction procedure referred in section 4.4?
Thanks so much!
Thanks for sharing this code and your paper, it is very helpful!
I'm starting a project where I would like to predict the brainage using my own training data (different kind of subjects). Since your results seem very promising, I would like to use similar architecture. Could you possibly also share how your trained the neural network and which code files I would need for this?
Hi,
This toolbox is wonderful and really fast . Would you please provide a cpu version of it ?
The pretrained model is fit into UK biobank of which age distribution is [42,64]. But my dataset age distribution is [20,86].
If I use it for validation, would it be not working well on my dataset?
I have tried to fine-tune pretrained model, but your pretrained model outputs_dim=40, mine 70...it does not match...
What can I do?
Your work looks amazing; could you kindly upload a license file?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.