smutch / cov3d

A deep learning model to detect the presence and severity of COVID19 in patients from CT-scans.

cov3d

A deep learning model to detect the presence and severity of COVID19 in patients from CT-scans. It has been submitted as part of the workshop 'AI-enabled Medical Image Analysis – Digital Pathology & Radiology/COVID19 (MIA-COV19D)' at the European Conference on Computer Vision (ECCV) in 2022 and the IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP) in 2023.

Warning

This project is part of a computer vision competition and has not been released for clinical use.

The following is a quick introduction to running cov3d. For more detailed information, see the documentation and the accompanying papers.

Installation

Install cov3d with pip:

pip install git+https://github.com/rbturnbull/cov3d.git

Training

To train cov3d, use this command:

cov3d train --directory ./scans

The training data and the two CSV files are part of the COV19-CT-DB Database which is available from MIA-COV19D workshop team.

The default hyperparameters for training can be overridden using the command-line interface. All options for training the model can be seen with the command:

cov3d train --help

Also see the training options in the Command Line Interface documentation.

Inference

If you have a trained model, you can infer the presence and severity of COVID19 from a CT scan with the command:

cov3d infer --pretrained <path/to/model> --scan <path/to/ct-scan>

All options for inference can be found with the command:

cov3d infer --help

Also see the inference options in the Command Line Interface documentation.

Credits

• Robert Turnbull <[e-mail]>
• Created using torchapp (https://github.com/rbturnbull/torchapp)