Mass General Brigham tackles network upgrade, AI pilots

Shoring up the foundation

Mass General Brigham has two academic medical centers, three specialty hospitals and 12 community hospitals across New England. Supporting that ecosystem are approximately 5,000 doctors, 39,000 nurses, 15,000 researchers and 85,000 employees.

But the hospital's network design was outdated, and its devices needed a refresh. One of the priorities for Sriraman was modernizing the network with software-defined networking from Cisco to better support end-to-end network performance throughout the hospital ecosystem.

"If your robotic surgical machine in the [operating room] gets disconnected from the network, you're stopping surgery in the middle," Sriraman said. "Everything we do in the hospital is extremely connected."

In addition, Sriraman said his team is investing in the following areas:

• Updating and patching the hospital's devices to the latest versions.
• Replacing endpoint devices, such as desktops and laptops.
• Investing in data analytics for asset management and research.

For example, asset management helps the hospital predict the time frame of a patient moving from the emergency room to a hospital bed, while data analytics gives hospital researchers access to data they can use to cure diseases.

Evaluating AI's value

Mass General Brigham is known as an innovative hospital ecosystem, Sriraman said, so it's unsurprising that AI is an important initiative. But the hospital takes a careful approach to implementing AI, following its tagline of using AI and generative AI (GenAI) responsibly and appropriately.

"Not every problem can be solved by GenAI or AI," he said. As a result, Sriraman said he and his team take the following steps when considering where to apply AI:

1. Evaluate appropriate use cases for AI.
2. Assess performance, and ensure ongoing monitoring of the AI tool.
3. Build the technology infrastructure to support GenAI.
4. Ensure governance to define and adjudicate the responsible use of AI.

The hospital has launched AI pilots to take notes during doctor-patient encounters, read CT scans, build data models and reduce research time. In one use case, hospital researchers use AI to parse through patients' postpartum analysis and detect possible PTSD symptoms. Sriraman said they provided a secure system where researchers can enter sensitive personal health information into ChatGPT without OpenAI using that PHI for prompt engineering and training.

Sriraman acknowledged the buzz, excitement and anxiety around AI. But he stressed the importance of starting with small use cases and evaluating AI performance before going all in.

"Everything we do has an impact on patients, so we want to measure the performance of AI before we scale," Sriraman said.

Jennifer English is executive editor for TechTarget's Networking and Cloud sites. She joined TechTarget as a writer and editor after graduating from the University of New Hampshire in 2016.