responsive site templates

AI-CT-COVID-19

A collaborative online AI engine for precise CT-based COVID-19 diagnosis


We deliver an AI-based CT diagnostic tool based on the concept of federated learning to provide people worldwide an effective AI model for precise CT-COVID diagnosis.

  • Federated learning - A decentralized architecture where the AI model is distributed to, and executed at, each host institution without the need for centralization, thus removing entirely the risk of data exposure.
  • Initial global model - 3D CNN architecture of light 3D-DenseNet, capable of classifying CT cases into health, COVID-19, bacterial pneumonand and other viral pneumonia. 
  • Online diagnostic tool- A tool providing access for users to diagnose the CT case with the global model we owned.

Federated Learning Framework

The strategy to preserve privacy by design enables medical centers to collaborate on the development of models, without the need of directly sharing sensitive clinical data with other insitutions.

APIs

APIs for the implementation of federated learning.

Download Initial Model

Download model architecture and weight of the global model from the server.

Train Locally

Modify configuration and start training with local data.

Upload Parameters

Uploading model weight with encryption to update the global model.

Mobirise

Online Diagnostic Tool 

1

Upload CT Images

Upload your CT case to the online diagnostic tool. (.DICOM format)

2

Start Diagnosis

Hit "Start Diagnosis" to start.

3

Get Results

Return the diagnostic result and provide the access to download the report.

Collaboration

Mobirise



This website and the reported AI engine based on federated learning framework support UN Sustainability Development Goals (UN SDGs), especially #3 Good Health and Well-Being.

- The team

Copyright © Huazhong University of Science and Technology. All rights reserved.
Contact ai.ct.covid.19@gmail.com for the international collaboration of AI engine.