Add ADNI dataset and Alzheimer's Disease classification pipeline

[laubryan]

Contributor

• Name: Bryan Lau
• NetID (Email): bryan16 (bryan16@illinois.edu)

Contribution

This PR is a full pipeline implementation of the Alzheimer's Disease classification model described by Liu et al., consisting of dataset, task and model for ADNI MRI images dataset.

Paper

This work is an implementation of the methods and ideas presented in:

• Liu et al., On the Design of Convolutional Neural Networks for Automatic Detection of Alzheimer's Disease
• arXiv: https://arxiv.org/abs/1911.03740

Description

This PR adds support for ADNI MRI brain scan images using a PyHealth-compatible dataset and processing via a compatible task and multi-class classification model, based on the paper "On the Design of Convolutional Neural Networks for Automatic Detection of Alzheimer's Disease" by Liu et al.

Changes

• Implemented PyHealth-compatible ADNIDataset for ADNI MRI images, with YAML configuration file
• Added NIfTIImageProcessor for NIftI MRI images
• Added AlzheimersDiseaseClassification task
• Implemented AlzheimersDiseaseCNN 3D CNN model for classification
• Added unit tests
• Added example scripts and notebook
• Added new RST files for pipeline components
• Updated related RST documentation indices

Files to Review

Category	Files
Dataset	pyhealth/datasets/adni.py pyhealth/datasets/configs/adni.yaml
Processor	pyhealth/processors/nifti_image_processor.py
Task	pyhealth/tasks/ad_classification.py
Model	pyhealth/models/alzheimers_cnn.py
Tests	pyhealth/tests/core/test_adni.py
Examples	pyhealth/examples/adni_adclassification_cnn.ipynb pyhealth/examples/adni_ad_classification.py pyhealth/examples/adni_ad_synthetic_data.py
Documentation	pyhealth/docs/api/datasets.rst pyhealth/docs/api/datasets/pyhealth.datasets.ADNIDataset.rst pyhealth/docs/api/processors.rst pyhealth/docs/api/processors/pyhealth.processors.NIftIImageProcessor.rst pyhealth/docs/api/tasks.rst pyhealth/docs/api/tasks/pyhealth.tasks.AlzheimersDiseaseClassification.rst pyhealth/docs/api/models.rst pyhealth/docs/api/models/pyhealth.models.AlzheimersDiseaseCNN.rst

Notes

• Access to download ADNI files requires approval through IDA (https://adni.loni.usc.edu/data-samples/adni-data/#AccessData)
• Example scripts and unit tests generate and use synthetic data