Skip to content

Open Source project to help predict Pulmonary Fibrosis from CT scans and clinical data.

License

Notifications You must be signed in to change notification settings

darwinai/FibrosisNet

Repository files navigation

Fibrosis-Net

FibrosisNet Logo

Note: The Fibrosis-Net model provided here is intended as a reference model that can be built upon and enhanced as new data becomes available. It is currently at a research stage and not a production-ready model (not meant for direct clinical usage). We are working continuously to improve it as new data becomes available. Please do not use Fibrosis-Net for self-diagnosis and seek help from your local health authorities.

image classification results with critical factors highlighted by GSInquire
Example images and their associated critical factors (highlighted in white) as identified by GSInquire.

Pulmonary fibrosis is a devastating chronic lung disease that causes irreparable lung tissue scarring and damage, resulting in progressive loss in lung capacity; it has no known cure. A critical step in the treatment and management of pulmonary fibrosis is the assessment of lung function decline. Computed tomography (CT) imaging is a particularly effective method for determining the extent of lung damage caused by pulmonary fibrosis. Motivated by this, we introduce Fibrosis-Net, a deep convolutional neural network design tailored for the prediction of pulmonary fibrosis progression from chest CT images. More specifically, we leveraged machine-driven design exploration to determine a strong architectural design for CT lung analysis, upon which we built a customized network design tailored for predicting forced vital capacity (FVC) based on a patient's CT scan, initial spirometry measurement, and clinical metadata. Finally, we leveraged an explainability-driven performance validation strategy to study the decision-making behaviour of Fibrosis-Net and verify that predictions are based on relevant visual indicators in CT images. Experiments using the OSIC Pulmonary Fibrosis Progression Challenge benchmark dataset showed that the proposed Fibrosis-Net is able to achieve a significantly higher modified Laplace Log Likelihood score than the winning solutions on the challenge leaderboard. Furthermore, explainability-driven performance validation demonstrated that the proposed Fibrosis-Net exhibits correct decision-making behaviour by leveraging clinically-relevant visual indicators in CT images when making predictions on pulmonary fibrosis progress. Fibrosis-Net is available to the general public in an open-source and open access manner as part of the OpenMedAI initiative. While Fibrosis-Net is not yet a production-ready clinical assessment solution, we hope that releasing the model will encourage researchers, clinicians, and citizen data scientists alike to leverage and build upon them.

For a detailed description of the methodology behind Fibrosis-Net and more information, please click here.

If you find our work useful, you can cite our paper using:

@misc{wong2021fibrosisnet,
      title={Fibrosis-Net: A Tailored Deep Convolutional Neural Network Design for Prediction of Pulmonary Fibrosis Progression from Chest CT Images}, 
      author={Alexander Wong and Jack Lu and Adam Dorfman and Paul McInnis and Mahmoud Famouri and Daniel Manary and James Ren Hou Lee and Michael Lynch},
      year={2021},
      eprint={2103.04008},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Files

  • Fibrosis-Net is comprised of two main components, FibrosisNet-CT and FibrosisNet-Clinical. The FibrosisNet-CT component takes in the provided CT scans along with clinical metadata, while the FibrosisNet-Clinical component takes in the provided clinical metadata.
  • A pre-trained model, FibrosisNetCT.pb, is included in the "models/" directory.

How to Reproduce Our Results

To reproduce our results for Fibrosis-Net, first clone this repository, and download the dataset here.

After extraction of the code and dataset, navigate into the main folder and install dependencies using:

pip3 install -r requirements.txt

Now you can run the following code to generate our Kaggle submission file. Make sure to modify the arguments as necessary. Running the code using CT_WEIGHT=1.0 achieves -6.8188 private score. Running it with CT_WEIGHT=0.96 achieves -6.8195 private score. Note that the DarwinAI team have reported a 0.0001 deviation in score between different Kaggle accounts and days of testing.

python3 kaggle_submission.py --datapath="osic-pulmonary-fibrosis-progression/" --modelpath="models/" --ctweight=1.0 --outputfile="submission.csv"

Requirements

The main requirements are listed below. A full list can be found in "requirements.txt".

  • Tested with Tensorflow 1.15
  • OpenCV 4.5.1
  • Python 3.6
  • Numpy 1.20.0
  • Pandas 1.2.1

Using the evaluation script

In order to perform custom evaluation using the evaluation script, you must first run the "kaggle_submission.py" script (this automatically generates the additional models required for evaluation).

Set up the evaluation inputs. Refer to the directory "example_input" for an example. There must be a .csv file inside, with these columns:

  • Patient
  • Weeks
  • FVC
  • Percent
  • Age
  • Sex
  • SmokingStatus
  • PredictWeek

For each patient listed in the .csv file, a folder containing CT scan images should be available in the same directory. Each folder should correspond to a patient in the .csv file, linked by folder names.

Run the following code and make sure to modify the arguments accordingly. If successful, a .csv file will be generated, containing all of the predictions for 100 weeks for each patient.

python3 fibrosisnet_eval.py --modelpath="models/" --inputpath="example_input/" --ctweight=1.0 --outputfile="results.csv"

About

Open Source project to help predict Pulmonary Fibrosis from CT scans and clinical data.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages