CoCalc provides the best real-time collaborative environment for Jupyter Notebooks, LaTeX documents, and SageMath, scalable from individual users to large groups and classes!
CoCalc provides the best real-time collaborative environment for Jupyter Notebooks, LaTeX documents, and SageMath, scalable from individual users to large groups and classes!
Path: blob/main/seminar6/3D_fmriICA_network.ipynb
Views: 63
Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Collecting monai
Downloading monai-1.0.0-202209161346-py3-none-any.whl (1.1 MB)
|████████████████████████████████| 1.1 MB 4.9 MB/s
Requirement already satisfied: torch>=1.7 in /usr/local/lib/python3.7/dist-packages (from monai) (1.12.1+cu113)
Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.7/dist-packages (from monai) (1.21.6)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from torch>=1.7->monai) (4.1.1)
Installing collected packages: monai
Successfully installed monai-1.0.0
Seminar 6.2 : Deep Learning methods for fMRI data
In the seminar we try to implement 3D CNN+GRU model on ICA fMRI data to classify Schizophrenia vs Control on SchizConnect dataset.
fMRI data is 4D tensor with 3 spatial dimensions and one temporal. In last part of seminar we selected 30 ICA components from fMRI data.
Data from SchizConnect was preproccesed by fmriprep
Useful links:
Monai docs: https://docs.monai.io/en/latest/api.html
SchizConnect: http://schizconnect.org/
fMRIPrep : https://fmriprep.org/en/stable/index.html
Firstly, we need to download preprecessed fMRI data:
1. Load Nifty fMRI data with MONAI Datasets
2. Create 3D CNN+GRU for classify Schizophrenia vs Control
TO DO: Try to improve the quality of the classifacation ( Schizophrenia VS Control)
Try ResidualUnit Blocks, LSTM block , different numbers of blocks
Add schedulers + hyperparameter optimization (lr, weight decay)
Add augmentations from MONAI
Write report with comparison of all methods +code
Second part:
Extract ROI from fMRI, build connectivity matrices
Train 2D CNN model from MONAI on connectivity matrices
Write report with comparison of all methods +code