https://github.com/basedrhys/non-roi-masking

Evaluation of non-ROI masking to improve OOD generalization in chest x-ray disease classification
https://github.com/basedrhys/non-roi-masking

chest-xrays classification medical-data medical-imaging ood-generalization

Last synced: about 1 month ago
JSON representation

Evaluation of non-ROI masking to improve OOD generalization in chest x-ray disease classification

• Host: GitHub
• URL: https://github.com/basedrhys/non-roi-masking
• Owner: basedrhys
• License: apache-2.0
• Created: 2021-11-20T06:25:34.000Z (over 3 years ago)
• Default Branch: master
• Last Pushed: 2022-09-30T19:46:36.000Z (over 2 years ago)
• Last Synced: 2025-02-15T19:48:42.572Z (3 months ago)
• Topics: chest-xrays, classification, medical-data, medical-imaging, ood-generalization
• Language: Jupyter Notebook
• Homepage:
• Size: 10.7 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# An Emperical Evaluation of Reducing Spurious Signals in Chest X-Rays via Non-ROI Masking

[Weights and Biases Project](https://wandb.ai/basedrhys/ml4h)

[Full Report](https://drive.google.com/file/d/1XB_jxtC2HqRp0ic8NvW7ChnIDJYAwKkv/view?usp=sharing)

![Graph](graph.png)

### Contents

* `0-seg_train.ipynb` - Train the lung segmentation model

* `1-seg_apply.py` - Create masks for a classification dataset via the segmentation model

* `2-smooth_masks.py` - Postprocess the predicted lung masks to smooth them out

* `3-create_chexpert.ipynb` - Create the binary classification version of CheXpert

* `3-create_datasets.py` - Apply the smoothed masks and create train/val/test splits of classification dataset

* `4-clf_train.ipynb` - Train the downstream classification model

* `4-clf_train.py` - Train the downstream classification model

* `5-eval.py` - Evaluate the trained classification models