DISCLAIMER: This a RESEARCH USE ONLY tool. Its purpose is to help clinicians and researchers explore the value of AI in medical imaging. These AI models are not regulated or validated and only for suggestion purposes.
Please note that the Arterys Marketplace is in beta. The imaging data you upload will not be retained past the beta program.

COVID-19 lesion segmentation

Pingan Coronavirus Medical Imaging AI System aims to assist doctors to diagnose more effectively and efficiently by leveraging AI and could help to control the epidemic through early diagnoses and treatment.

Your own data

Cost: Free during Beta

Pingan Coronavirus Medical Imaging AI System can generate smart analysis results in around 15 seconds for normal CT series scanning, the system has four major fuctions:

1) Lesion segmentation for generate quantitative measurement results
2) Lung segmentation for quantification(percentage of lesion)
3) Lobe segmentation for lesion location
4) Image registration for follow up

And can be used in the following scenarios:

1) Coronavirus Suspected Case Screening and Early Detection
2) Coronavirus Suspected Case Quantitative Analysis and Reporting
3) Follow-up and Treatment Effectiveness Evaluation.

Intended Use

1) DICOM CT Series with thickness no more than 5mm
2) bodypart must be CHEST and should be in axial file


This model is designed to detect and segment pneumonia lesions, not only for covid 19 but also opacities in other pneumonia diseases, and quantitatively analyze those lesions.

Information on training data

The model is trained on data from several sources in China in an end-to-end manner.

Model performance metrics

For classifying negative VS positive
Accuracy: 0.95
Sensitivity: 0.95
Specificity: 0.94
For lesion segmentation
DICE: 0.834


PingAn Technology Artificial Intelligence Center