A Rapid, Accurate and Machine-agnostic Segmentation and Quantification Method for CT-scan-based COVID-19 Diagnostics

Overview

This repository provides the predictive model described in the paper:

Longxi Zhou, et al. "A Rapid, Accurate and Machine-agnostic Segmentation and Quantification Method for CT-based COVID-19 Diagnosis"

Contents

• 01.introductory.demo contains and example prediction of our model
• 02.our.model contains our full-fledged model
• 03.baselines.demo contains code for baselines

The trained models of our model and the baseline methods are stored on Google Drive. Please respect the folder structure in the drive when downloading.

Example Data on the Google Drive:

The data for 02.our.model is in in 02.our.model/patients/. Our method will preprocessing these files, predict, and visualize the infection segmentations. The data for models in 03.baselines.demo is in CT_scan_spatial_signal_normalized/, which are same arrays with arrays stored in ./02.our.model/standard/patient_id/time_point/ after the preprocessing. Read the readme files for these comparisions for detailed information. The Lung_segmentation_mask/ stores the lung_masks for the scans: 1 means inside lungs, 0 means outside lungs. All methods used the same lung masks to exclude false-positives when we did the quatitative analysis.

Contact

If you request our training code/simulation model for COVID-19, please contact Prof. Xin Gao at [email protected].

test