A Comprehensive Use Case of Clinical Deployment of AI Models In Hospitals
, Mass General Brigham
, NVIDIA
With the rapid growth of AI algorithms, hospitals can benefit from new techniques, especially during the COVID-19 pandemic. However, integrating AI models into complex clinical workflows to make a real impact isn't easy. We'll discuss how to deploy an AI algorithm into an end-to-end AI hospital-deployment architecture in a short time. We'll review the infrastructure design, Clara pipeline definitions, and connection between each component. Our results demonstrate that decomposing overall architecture into modular components increases the efficacy, flexibility, and reuse in the actual deployment. Starting with a COVID x-ray AI classification model, we deployed it in 15 days and enabled 450-600 COVID AI-inferred studies per day.