COVID19 CT Scan Segmentation
[
]()
GOAL: Identifying infections in CT scan images of COVID19 patients using CNNs
COVID-19 patients usually develop pneumonia which rapidly progress to respiratory failure. Computed Tomography (CT) scan images play a supporative role in rapid diagnosis and the severity of the disease. Models that can find evidence of COVID-19 and/or characterize its findings can play a crucial role in optimizing diagnosis and treatment, especially in areas with a shortage of expert radiologists. This repository contains a package to identify lungs infections in CT scan images.
To download training and testing datasets see this
Installation
The package is written in Python and can be installed using python setup.py install or pip install . The you should be able to use the packaage:
from ctseg import patient
patient_data = patient.PatientData("testing-set/patient00")PatientData loads CT images which are used to train a convolutional network. Find out more details see how2useit/how2use-patient.py.
Using segmented lungs, CT image were cropped and fed into a convolutional network to train segmenting for infections (see ctseg/models/convnet.py). Models were trained and the outputs, including weights, were saved in the outputs folder. To explore the model performance see scripts/evaluate.py
To test the model on new patient data, fill out testdir in the defaults.config file and run
python -u scripts/run_on_test_patients.py defaults.config
About this dataset
This dataset contains images from 61 patients (divided into 55 training and 6 testing) diagnosed with COVID-19 (see references). Files containing corresponding segmentations of lungs and infections made by experts are included as well.
Acknowledgements
[1] - Paiva, O., 2020. CORONACASES.ORG - Helping Radiologists To Help People In More Than 100 Countries! Coronavirus Cases. See here
[2] - Glick, Y., 2020. Viewing Playlist: COVID-19 Pneumonia. Radiopaedia.Org. Available here
[3] - Kaggle dataset available here
[4] - Zhang, Kang, Xiaohong Liu, Jun Shen, Zhihuan Li, Ye Sang, Xingwang Wu, Yunfei Zha et al. "Clinically applicable AI system for accurate diagnosis, quantitative measurements, and prognosis of covid-19 pneumonia using computed tomography." Cell (2020).