UC San Diego uses AWS cloud to deploy homegrown AI algorithms for COVID-19

Dr. Albert Hsiao is an associate professor of radiology at the University of California San Diego School of Medicine and a radiologist at UC San Diego Health.


His laboratory has been developing algorithms for X-ray, CT and MRI scans for a few years, and the primary problem has been the ability to deploy these artificial intelligence algorithms and scale compute resources for use in the clinic.


“We previously considered purchasing hardware to satisfy compute resources we needed to bring the technology to life, but this required extensive security reviews and uncertainty on who would ultimately manage equipment failures,” Hsiao explained.

“When Amazon Web Services offered to provide resources to hospitals to address the COVID-19 pandemic, it seemed like a good mechanism to bring some of our AI technologies into the clinic.”

So UC San Diego Health engineered a new machine-learning method to help diagnose pneumonia earlier – a condition associated with severe COVID-19 cases. This detection allows doctors to quickly triage patients to the appropriate level of care, even before a COVID-19 diagnosis is confirmed.


There are a variety of major cloud computing vendors on the market today. Some of these include Google Cloud, Microsoft Azure and Salesforce.


“We were able to connect our hospital imaging informatics infrastructure with AWS to run our AI algorithms in the cloud, connecting with our hospital PACS system,” Hsiao said.


UC San Diego Health has been able to run roughly 10,000 chest X-rays through its algorithm thus far, and identified several patients with COVID-19 pneumonia on chest X-rays who otherwise may not have been diagnosed.

The AI algorithm provides support for radiologists, and emergency room and other physicians to detect subtle but important findings that may be missed.

Early COVID-19 pneumonia is a perfect example of a disease that can be subtle to the inexperienced eye, but obvious to a subspecialist radiologist, Hsiao explained. The AI, the provider organization has learned, helps to bridge that gap.


“Certainly, to leverage AWS requires a lot of technical skills and experience,” said Hsiao. “It requires the collaboration of a multidisciplinary team to bring new technologies to life.

“This included radiologists, engineers, data scientists, IT staff and support of our hospital administration to accomplish this,” he added. “Like most things worth doing, it sounds impossible when you first start, but when it is so valuable, can quickly become your standard of practice.”