How SRI’s AI technology is changing the way healthcare is delivered


Governments around the world have encouraged healthcare providers to utilize Telehealth to safely care for patients

Eric Yeh, an Advanced Computer Scientist at SRI International, understands the profound impact that artificial intelligence (AI) is having on healthcare. Yeh has many years of experience applying and adapting AI to solve real-world problems. He continues this groundbreaking work at SRI International’s Artificial Intelligence Center.

One such project is Decoded Health, a spin-off of SRI Ventures, the corporate venture arm of SRI International. Decoded Health has developed a robotic processing automation (RPA) platform that uses AI and natural language processing (NLP) to improve the quality of care while helping physicians connect with more patients. NLP has been widely used in healthcare to transcribe and analyze notes and patient interactions.

Decoded Health, however, takes this even further. First, patients can communicate symptoms in their own words, a unique capability in a domain where most interaction is limited to checkboxes and one or two-word answers. Decoded Health combines NLP with complex machine learning to enhance natural language understanding (NLU) of patient expressions — instead of merely chest pain, a patient can state “I feel like there is an elephant sitting on my chest” — to capture the nuances in patient speech that human clinicians pick up to make an accurate diagnosis. In addition, the RPA platform generates personalized recommendations based on patient-reported symptoms, medical history from the electronic health record, and diagnostic and clinical criteria from clinical practice guidelines. Decoded Health’s comprehensive approach increases both the accuracy and efficiency of primary healthcare.

Transforming NLP for the marketplace

Yeh explains that Decoded Health began as a very narrowly focused, non-healthcare application. This application, which was called ARSENAL, had a very specific purpose: to process and analyze text from lengthy hardware documentation, which Yeh refers to as “some of the driest material you will ever read.” However, as Yeh explains, these documents contain critical information, such as ensuring a microwave oven door is closed while the oven is in use. The original ARSENAL application used NLP to extract the most crucial information from a large body of text.

“The purpose of NLP is to understand anything derived from human language, written or spoken. We were approached by SRI Ventures to adapt this original model for clinical purposes,” Yeh said. The end result is Decoded Health, the latest in a long history of SRI innovations in speech recognition.

Decoded Health integrates ARSENAL with other SRI AI and NLP technologies, such as analytics that look for emotional clues when a patient is speaking. This can provide the physician with insight into a patient’s mental state as part of the overall clinical assessment, leading to better-informed treatment plans. The power of Decoded Health lies in its ability to accurately gather and interpret information, improving the quality of care while allowing physicians to see more patients than ever via this novel platform.

Advanced telehealth at the right time

While the use of telehealth has significantly increased as a result of the coronavirus pandemic, its widescale adoption has demonstrated how useful and effective it can be. Telehealth is here to stay, and governments around the world have encouraged healthcare providers to use it to safely care for patients. The development of Decoded Health began long before the pandemic, and the platform has proven to be particularly useful when urgent care is needed. While many COVID-19 symptoms are similar to other ailments, including heart attacks and pneumonia, Decoded Health uses the patient’s natural language expression of their symptoms to help doctors ask better questions and differentiate between illnesses.

Importantly, for a nation with a physician shortage prior to the pandemic, the step-function increases in efficiency created by the Decoded Health platform will allow physicians to meet the increased demand for care in a time when the public sorely needs ready access to a doctor.

Collectively, Yeh and the Decoded Health team have more than six decades of experience in machine learning, health IT (information technology) and clinical healthcare. Yeh says he enjoys being at SRI because “it is a dynamic environment and one of the few places I’ve seen where a person can actually make a project happen. Everyone at SRI is working on such cool stuff. It feels like a new job every two or three years, and that is really exciting.”

Indeed, SRI International and its SRI Ventures arm both have a well-known reputation for breakthrough products. In addition to Decoded Health, SRI technologies are used in other healthcare applications, such as Curie AI. Curie AI is another SRI Ventures healthcare spin-off that developed a transformative AI for objectively quantifying respiratory distress to compute “a new fifth vital sign.” The technology is transforming the standard of care for proactive interventions in high-risk patients with chronic/acute respiratory conditions. Curie’s solution enables early detection of deterioration in the condition (2–4 days in advance as compared to the standard vitals measurement alone) driving life-saving outcomes, preventing adverse events like hospitalizations, and helping patients recover faster.

As of 2019, SRI Ventures has launched more than 50 spin-offs like Decoded Health and Curie AI, exceeding a $20 billion market cap.


1. Statista.

Read more from SRI