Ensembles of Generative Adversarial Networks as a Data Augmentation Technique for Alzheimer research. Machine Learning Engineer Nanodegree at Udacity - Capstone Project

Data scarcity is a regular problem in research, and in medicine it’s especially difficult to find datasets publicly available. The main reason is its rarity, by definition images of anomalies are scattered and/or not common, and also there are a lot of legal issues that prevent sharing personal information about patients. In this talk, I’m going to give an introduction to ensembles, and more specifically to Generative Adversarial Networks (GANs) ensembles, applied to this problem of data scarcity, generating MRI images of demented brains that could be use in Alzheimer research.

• Presented at PyCon Italia 2023 - https://pycon.it/en/event/ensembles-of-gans-as-a-data-augmentation-technique-for-alzheimer-research

• Deck: https://speakerdeck.com/p1nox/ensembles-of-gans-as-a-data-augmentation-technique-for-alzheimer-research

• Video: https://youtu.be/6M0lUv9poaw?si=Kffsks6HrxZH267w

• Presented at PyCon Bolivia 2022: https://bo.pycon.org

• Deck: https://speakerdeck.com/p1nox/ensemble-of-gans-as-a-data-augmentation-technique-for-alzheimer-research

# install dependencies
conda create -n gan_ensembles python=3.6.5
conda activate gan_ensembles
pip install -r requirements.txt
conda install pytorch torchvision torchaudio cudatoolkit=11.1 -c pytorch -c nvidia
# start jupyter server
jupyter notebook

Notebooks in this project were designed to be executed in sequence:

1. Download MRIs dataset - http://localhost:8888/notebooks/download_dataset.ipynb
2. Download MRIs dataset - http://localhost:8888/notebooks/data_exploration_viz.ipynb
3. Data preprocessing and Data splitting - http://localhost:8888/notebooks/data_preproc_split.ipynb
4. DCGAN implementation and Control Model (CM) training - http://localhost:8888/notebooks/dcgan_control_model.ipynb
5. DCGAN Ensemble Model 1 (eGANs) training - http://localhost:8888/notebooks/dcgan_ensemble_model_1.ipynb
6. DCGAN Ensemble Model 2 (seGANs) training - http://localhost:8888/notebooks/dcgan_ensemble_model_2.ipynb
7. DCGAN Ensemble Model 3 (cGANs) training - http://localhost:8888/notebooks/dcgan_ensemble_model_3.ipynb
8. Metrics visualisation and conclusion - http://localhost:8888/notebooks/metrics_viz_outro.ipynb

Original proposal, corrected proposal, and final project report included in capstone_project_docs folder.