Prediction of misfolded proteins spreading in Alzheimer’s disease using machine learning

Authors

Luca Gherardini
Alessandro Crimi
Aleksandra Zajdel (Pestka)

Simulations and models used for predicting th progression of Alzheimer disease:

• Constrained Multivariate Autoregressive Model (MAR)
• Network Diffusion Model (NDM)
• Epidemic Spreading Model (ESM)

Citing this work

If you use this code, please cite this paper: "Prediction of misfolded Proteins using Machine Learning]{Prediction of misfolded Proteins spreading in Alzheimer's disease using Machine Learning and Spreading Models" Gherardini, Zajdel, Pini, Crimi, Cerebral Cortex 2023 https://academic.oup.com/cercor/advance-article/doi/10.1093/cercor/bhad380/7311320

Workflow

1. Install dependencies.
   To install the necessary dependencies you can use pip install -r requirements.txt. It is advised to use it within a python3 virtual environment with Python3.9.
2. Please, format your data according to the BIDS format:

├── ADNI
   └── derivatives
       ├── sub-AD4009
       │   ├── ses-baseline
       │   │   ├── dwi
       │   │   ├── anat
       │   │   └── pet
       │   └── ses-followup
       │       └── pet
       └── sub-AD4215
           ...

3. Perform preprocessing on your dataset

cd src/preprocessing
python3 main.py <img_type> <dataset_path> <cores> <atlas_file>

img_type: type of image to look up for (dwi, anat or pet) dataset_path: absolute or relative path of the input folder containing images (it can be '.' if its in the same path). Please note that the pipeline will produce a directory named 'derivatives' inside the specified path. cores: the number of cores to use (-1 uses all available). atlas_file: the path for the atlas to use during registration step. Insert 'anat' to coregister the current image to the corresponding anatomical T1-weighted file (if available). The CLI options are facultative and will be asked before starting the preprocessing if not specified.

4. Generate tractography and connectivity matrix from DWI data.

cd src/tractography
python3 TRK_CM_generation.py

5. Extract average MP concentration in brain regions.

cd src/analysis
python3 extract_regions_means_PET.py

6. Prepare dataset for training and testing.

cd src/dataset_preparing
python3 create_dataset.py <cores> <threshold>

threshold: the tolerance (between 0 and 1) with which discarding a subject if the Amyloid-beta concentration at baseline is >= threshold * followup 7. Choose model for making predicitons (including training and testing):

cd src/simulations
python3 simulation_MAR.py <category> <cores> <train_size> <lamda> <matrix> <use_binary> <iter_max> <N_fold>

category: clinical group on which perform simulations. A file in the dataset directory named dataset_.json must exhist. lambda: parameter in [0,1] used to set an L2 regression on the gradient descent optimization. Its utilization has not been considered in the papers results. matrix: prior to use for matrix A (0: CM, 1: diag(baseline), 2: random, 3: identity) use_binary: use binary matrix as constraint (1) or no (0)

python3 simulation_ESM.py <category> <cores> <beta_0> <delta_0> <mu_noise> <sigma_noise>
python3 simulation_NDM.py <category> <cores> <beta>
python3 simulation_GCN.py <category> <matrix> <epochs>
python3 simulation_CDRMR.py <category> <cores> <train_size> <iter_max> <N_fold>

Project Organization

├── README.md                                      
├── requirements.txt                # requirement packages
├── config.yaml                     # configuration file for tractography and connectome generation
├── publications                    # reference publications
├── data
│   └── atlas                       # brain atlas
└── src                             # source code 
    ├── analysis                   
    │   ├── extract_regions_means_PET.py # extracting average MP concentration from brain regions
    │   ├── prediction_analysis.py
    │   ├── prediction_visualization.py
    │   ├── statistics_connectomes.py
    │   └── statistics_PET.py
    ├── dataset_preparing           # dataset preparation: input and output for MAR and simulations 
    │   ├── create_dataset.py
    │   ├── dataset_av45.json
    ├── preprocessing               # preprocessing scripts 
    │   ├── main.py
    │   └── utils
    │       ├── brain_extraction.py
    │       ├── brain_segmentation.py
    │       ├── cerebellum_normalization.py
    │       ├── denoising.py
    │       ├── eddy_correction.py
    │       ├── flatten.py
    │       ├── gibbs.py
    │       ├── motion_correction.py
    │       └── registration.py
    ├── simulations                 # simulation scripts 
    │   ├── simulation_NDM.py 
    │   ├── simulation_ESM.py
    │   ├── simulation_MAR.py
    │   ├── simulation_GCN.py
    │   ├── simulation_CDRMR.py
    │   ├── utils.py
    │   └── utils_vis.py
    └── tractography                # tractography and connectomes generation 
        ├── TRK_CM_generation.py
        └── utils.py