Bayesian Neural Networks for Uncertainty Estimation of Imaging Biomarkers

Accepted in MICCAI-MLMI-2020

More details here: MICCAI-MLMI-2020

Folder Structures:

• dataset_groups: It holds various datasets with their respective processing code in it.
• projects: It holds various bayesian architectures i.e. 4 that we used for our experiments. Fully Bayesian, Quicknat with dropout, Probabilistic U-Net, Hierarchical U-Net.
• interface: It has all the base class for solver, data processing pipeline, evaluator and run setup. Provides a consistent platform for all projects to train and evaluate.
• utils: To have utility functions like logger & notifier to mention a few.
• stat_analysis: This folder contains post segmentation data analysis with diesease classification and group analysis stuff using python and R toolkit.

if you like the paper, and willing to extend the work, please cite:

@inproceedings{senapati2020bayesian,
  title={Bayesian Neural Networks for Uncertainty Estimation of Imaging Biomarkers},
  author={Senapati, Jyotirmay and Roy, Abhijit Guha and P{\"o}lsterl, Sebastian and Gutmann, Daniel and Gatidis, Sergios and Schlett, Christopher and Peters, Anette and Bamberg, Fabian and Wachinger, Christian},
  booktitle={International Workshop on Machine Learning in Medical Imaging},
  pages={270--280},
  year={2020},
  organization={Springer}
}

Code Authors

• Jyotirmay Senapati - jyotirmay-senapati