FirasGit / chest_radiography_ai_vs_hi

This repository contains the code for our paper: Artificial Intelligence for Clinical Interpretation of Bedside Chest Radiographs

Installation

To clone this repository please run

[e-mail]:FirasGit/chest_radiography_classification.git

then navigate into the cloned repository using

cd chest_radiography_classification

and create a conda environment

conda create -n chest_radiography_classification python=3.8

Once activating the environment using

conda activate chest_radiography_classification

run

pip install -r requirements.txt

to install all dependencies.

Train the model

In order to train the model you need to request access to our internal dataset or use your own dataset. To train the model, run the following command

python training/lightning_trainer.py meta.prefix_name=<your_preferred_run_name> model.name=<model_name> optimizer.learning_rate=<learning_rate> annotations.path_to_train_annotation_csv=<path_to_training_set> annotations.path_to_valid_annotation_csv=<path_to_validation_set> annotations.path_to_test_annotation_csv=<path_to_test_set> optimizer.loss_fnc=<loss_function> meta.batch_size=<batch_size>

For a more detailed view over all configuration settings, navigate to configs/train_config/base_cfg.yaml