Learning is a lifetime pursuit for most people and required for most occupations. If you ask which occupations require the strictest nonstop learning, the two likely candidates are medical doctor and IT specialist. John Yu, Founder and Chief Executive Officer of Meridian Medical Network Corp., works with both.
Artificial intelligence (AI), now a household name across the world, is learnable IT — it has the ability to learn and therefore currently resides on the “highest food chain of information technology,” according to Yu.
“We want to put these two together to learn: doctor and AI,” says Yu. “The doctor needs to learn. AI needs to learn. If we put these two together, we could probably come up with [solutions to] the issues of health care infrastructure, health care transformation, and innovation.” Creating smart health care, or learnable health care, is a primary focus of Meridian Medical.
To facilitate the best learning for smart health care, it’s all about symbiosis. “Symbiosis is fundamental when building on learnings between the human and the machine, or the doctors and AI,” Yu says. He describes the symbiosis, or close, long-term interaction between two different biological organisms, between a rhinoceros and a bird — the bird sits on the rhinoceros and in exchange eats bothersome insects and parasites.
There are different ways of categorizing the human and machine symbiosis system. The first is a two-way system of doctors and AI. The second is a three-way system: doctors, AI, and patients. “We want to facilitate the interactions among those biological organisms — doctors, AI, and patients — so that a chemical factor can be arrived at in terms of learning,” Yu says.
Yu envisions a human-to-machine learning symbiosis system that includes the human learnings of doctors and the machine learnings of AI, and then when the two interact and operate with each other they provide smart health services for patients. In turn, patient-reported outcomes and other resulting patient health data is fed back to the doctor and the machine, creating the necessary conditions for ultimate learning.
“We have this doctor as executor of this task of providing some health services and the machine learning, AI, is the helping tool. Patients are the recipients of the services and also, importantly, the feedback agent,” he explains.
A deeper look at human and machine learning symbiosis reveals each side’s strengths and weaknesses, helping us to understand what to leverage and what to avoid. “A machine is fast, accurate, but naïve in terms of a lack of common sense, intuition — [it has] nothing,” Yu says. “A human? Slow, very inaccurate when processing large amounts of information, but brilliant — brilliant in common sense, intuition, and sympathy, which are all essential, all required to provide good, smart health care services.”
“Machine learning must be combined with human learning to leverage their strengths and avoid their weaknesses, respectively. So machine learning, in a fundamental level, must have humans in the loop. Likewise, human learning must have machines in the loop,” he explains. “In this way, we can work around the patient in the center, not only receiving the smart health services, but also having critical feedback to the human doctors and to the machine, AI.”
Yu describes an example of how Meridian actually did that, with the Jingde Experiment, the first large-scale field trial of primary care AI in China, using a smart device they developed called the Angel Robot.
The Angel Robot device is connected to the medical AI brain or cloud and serves as a convenient companion for primary care physicians (PCPs), helping them work on major tasks.
The first task for the Angel Robot is basic medicine, serving as a clinical decision support system (CDSS) for traditional Chinese medicine and Western medicine, helping doctors go through the entire diagnosis and treatment process. It not only provides a knowledgeable, authorized database, but it also acts as a dynamic engine, helping to lead and nurture PCPs at the grassroots level.
Next, under public health, Angel Robot helps with health management, family doctor services, chronic disease health assessments, and data collection, all of which are dynamic and can be automated — such as vital sign monitoring on the customer side — and tailored to each patient’s situation. Angel Robot also provides remote health consultation and appointment services under the telemedicine umbrella and AI-assisted functions such as facial recognition and voice access.
The Jingde Experiment, named for the county it took place in, paired Angel Robot with village “barefoot” doctors, many of them older adults, to provide care to a population of 150,000. Deployment of the smart tiered health care network from September 2017 to March 2019 resulted in 15,020 primary care services, close to 2,000 patient telemedicine services, and a 95% patient satisfaction rate, according to Yu.
The Jingde County Government called the smart health service “accessible, capable, and affordable,” and Yu says that when a delegation from the World Bank and World Health Organization visited, they confirmed that Angel Robot “could be adapted in similar settings in other places worldwide.”
Returning to the symbiosis between human and AI, Yu shares what PCPs have said about the helpfulness of Angel Robot: It “reduces the time for information input significantly,” “expands the types of diseases I can provide care for,” “is easy to operate,” and “is a great assistant for my daily work and should be scaled up.”
“The Angel Robot or AI and the PCP learning symbiosis work in this fundamental structure [across] the whole process of primary care services so that the AI learns from patients’ outcomes and from PCPs’ input as well, and PCP learning is assisted by AI,” explains Yu. “From the fundamental structure of this learning, the PCP and the Angel Robot must be formed with closed-loop feedback, must have continued, ongoing data collection, must have the patient’s outcomes and the PCP’s input and Angel Robot–assisted diagnosis, and then the treatments can be achieved from the PCP’s perspective.”
“This is what truly we call a new paradigm for smart health: AI and the doctor mutually learning and enhancing with each other, by give and take, respectively, working geared toward their strengths and [away from] their weaknesses, to put it all together,” says Yu. “In this way, we can make better AI, better doctors, and in the end, we want to make better patient outcomes.”
Quoting Dory in Finding Nemo, “just keep swimming,” Yu tells us likewise to “keep learning.”
“Let’s build up this human and machine symbiosis in whatever fashion to provide the necessary conditions and fertilize the ground for them to learn and work with each other. At the end, we’ll make better doctors, better AI, and better patient outcomes.”
From the NEJM Catalyst event China’s Changing Health Care: Global Lessons at Scale, held at Jiahui Health in Shanghai, April 25, 2019.