Convolutional Neural Networks for Classification of Alzheimer's Disease: Overview and Reproducible Evaluation [Models]

This file contains the pretrained models and the evaluation of the pipelines described in the paper Convolutional Neural Networks for Classification of Alzheimer’s Disease: Overview and Reproducible Evaluation.

Source code can be downloaded at: https://github.com/aramis-lab/AD-DL

Also, single files can be obtained at: https://aramislab.paris.inria.fr/clinicadl/files/models/v0.0.1/

The structure of the compressed file is as follows:

clinicadl_models/
├── 2D_slice
│   ├── baseline
│   │   ├── AD_CN
│   │   │   ├── best_model
│   │   │   └── performances
│   │   └── AD_CN_dataleakage
│   │       ├── best_model
│   │       └── performances
│   └── longitudinal
│       └── AD_CN
│           ├── best_model
│           └── performances
├── 3D_patch
│   ├── baseline
│   │   ├── AD_CN
│   │   │   ├── best_model
│   │   │   └── performances
│   │   └── sMCI_pMCI
│   │       ├── best_model
│   │       └── performances
│   └── longitudinal
│       ├── AD_CN
│       │   ├── best_model
│       │   └── performances
│       └── sMCI_pMCI
│           ├── best_model
│           └── performances
├── 3D_ROI_based
│   ├── baseline
│   │   ├── AD_CN
│   │   │   ├── best_model
│   │   │   └── performances
│   │   └── sMCI_pMCI
│   │       ├── best_model
│   │       └── performances
│   └── longitudinal
│       ├── AD_CN
│       │   ├── best_model
│       │   └── performances
│       └── sMCI_pMCI
│           ├── best_model
│           └── performances
├── 3D_subject
│   ├── baseline
│   │   ├── AD_CN
│   │   │   ├── best_model
│   │   │   └── performances
│   │   └── sMCI_pMCI
│   │       ├── best_model
│   │       └── performances
│   └── longitudinal
│       ├── AD_CN
│       │   ├── best_model
│       │   └── performances
│       └── sMCI_pMCI
│           ├── best_model
│           └── performances
├── autoencoders
│   ├── 3D_patch
│   │   ├── baseline
│   │   │   └── best_model
│   │   └── longitudinal
│   │       └── best_model
│   ├── 3D_ROI_based
│   │   ├── baseline
│   │   │   └── best_model
│   │   └── longitudinal
│   │       └── best_model
│   └── 3D_subject
│       └── baseline
│           ├── extensive
│           └── minimal
└── svm
    ├── baseline
    │   ├── AD_CN
    │   │   ├── all_subjects.tsv
    │   │   └── classifier
    │   └── sMCI_pMCI
    │       ├── all_subjects.tsv
    │       └── classifier
    └── longitudinal
        ├── AD_CN
        │   ├── all_subjects.tsv
        │   └── classifier
        └── sMCI_pMCI
            ├── all_subjects.tsv
            └── classifier

We provide the pretrained CNN models for the frameworks 3D subject-level, 3D ROI-based, 3D patch-level and 2D slice-level. This models can be found as a .pth.tar file (Pytorch format) inside the best_model folder for each framework (and for each fold). We also provide the autoencoders that initialize the training stage of the CNN networks. The performances folder contains the computed metrics for the correponding model (ACC, BA, etc). 

For the svn classification, we provide files with the dual coefficients, the support vector indices and the weights. Also, tsv files with the subject list.