NVIDIA’s AI Could Prove Helpful in COVID-19 Emergency

People critically infected with COVID-19 may soon have the odds, stacked in their favor. Nvidia and Massachusetts General Brigham Hospital have developed an AI model that can determine whether a person that enters an emergency room, requires supplemental oxygen hours, or even days after an initial exam.

Named CORISK, the original AI model was developed by scientist Dr. Quanzheng Li at Mass General Brigham. The system works by combining medical imaging with patient health records. In order to develop an effective AI mode, the team started an initiative called EXAM (EMR CXR AI Model). It’s the largest and most diverse federated learning initiative.

The system uses the NVIDIA Clara federated Learning Framework. Researchers at each hospital were able to use chest X-ray, patient virals and lab values to train a local model. The ultimate goal was to determine, as quickly as possible, whether the patient requires supplemental oxygen.

According to NVIDIA, Dr. Ittai Dayan, who leads the project, also co-led the EXAAM initiative with the company and facilitated to use of CORISK. Certain improvements were made by training the model on distributed data from a multinational, diverse dataset of patients across North and South America, Canada, Europe and Asia.

Each hospital’s AI model, uses a secure, in-house server for its data. But, a separate server hosted on AWS, houses the global deep neural network. NVIDIA says that each participating hospital gets a copy of the model to train on its own dataset. This negates the need to pool patient data into a single location, which in turn, protects patient privacy.

NVIDIA’s Clara could help immensely in the medical imaging field. By making said data, AI ready, healthcare facilities will have a 3rd eye, presiding over each diagnosis, helping to speed up the process and minimize mistakes. NVIDIA Clara’s AI-Assisted Annotation can provide APIs and a toolkit to bring AI-assisted annotation capabilities to any medical viewer.