When Dr. Fang Jin took part in a Beijing competition to diagnose brain tumors, he was determined to win. The prize was more than $160,000, to be used for research.
There were just two problems.
First, “the questions were very tough to answer,” he said.
Second, the competition was a robot.
By question two, he gave up trying to win. “My aim was just not to lose by too much.”
Billed as the first neuroimaging competition between humans and an artificial intelligence “doctor” named Biomind, the battle last week, televised on state TV, was designed to promote AI in medicine in China, a country where rural medical facilities lag behind city hospitals, which in turn are stressed and overloaded.
Biomind was developed in a joint venture between a Singaporean tech company, Hanalytics, and China’s prestigious Tiantan Hospital. Since December, when the venture started up, tens of thousands of medical images collected over a decade have been used to teach Biomind its job. After months of deep learning, the machine was ready for a competition against 25 experienced doctors at Beijing’s China National Convention Center testing their ability to analyze images of the brain.
Artificial intelligence is a key part of the government’s Made in China 2025 plan, which aims to make the nation a global leader in high-tech industries, including robotics, by that year. This worries the United States deeply: If China can lead the world in medical AI, could it rival U.S. hegemony in other, more threatening fields?
Testifying before a Senate committee in March, U.S. Trade Representative Robert Lighthizer spelled out fears about Chinese dominance in robotics and other fields such as advanced medical products, aeronautics, high-tech transportation and biopharma. “These are things that China listed and said we’re going to take technology, spend a couple hundred billion dollars and dominate the world,” said Lighthizer. “These are things that if China dominates the world, it’s bad for America.”
Concerns about China’s rise to rival U.S. technology, and American accusations that it has gained an unfair foothold by requiring U.S. firms to share technological expertise, are at the heart of Sino-U.S. trade tensions. China denies the accusations. Analysts talk about a tech cold war.
The AI competition had the glitz and drama of a TV game show, with flashing lights, heroic music and a glamorous presenter wearing a filmy white miniskirt and one dangly gold earring.
The event underscored how China’s vast access to medical data — with its 1.4 billion population and digitized data in all aspects of medical care — may give it an edge in medical applications of AI. China is encouraging its population to wear health monitors around the clock, offering scope for AI to mine huge amounts of information, with hopes that intelligent machines will be able to predict illnesses, or diagnose them earlier than doctors.
Every day I consult and diagnose over 100 patients, so ... I’d welcome having AI to improve my efficiency.
Before going up against Biomind, one competitor, Dr. Wang Chongqing, 54, said he did not think AI could replace doctors, but it could help perform some tasks.
“I think humans and machines are good at different things. They can complement each other,” he said. “I think that humans are better at diagnosing conditions in a more holistic way, taking into account gender, age and history and that helps to diagnose diseases better. Machines don’t have this ability yet.”
Another competitor, Dr. Zhong Di, 45, from Harbin, said she was not afraid of AI machines.
“I think this thing has to come one day and I’m ready for it. I think it will definitely help doctors. It might help to improve the efficiency and accuracy of the diagnosis,” she said.
In round one of the battle, Biomind and 15 doctors were given 15 questions and 30 minutes to examine brain images and provide the answer as to which tumors were shown, some extremely rare. In the second round, 10 more doctors competed with the machine on images relating to strokes. As the seconds ticked by, the pressure showed. One doctor puffed out his cheeks, looking perplexed, while another leaned close to her screen, eyes squinting.
Biomind, a boxy-looking black machine shaped something like a fat dumbbell, sat on a table with a cluster of experts fluttering about like waiters. Within 15 minutes, its little white light flashed as the machine finished the test. The doctors toiled on.
When the results came in, Biomind beat the doctors squarely in both rounds. In round one, it correctly answered 87% of the questions, versus 66% for the doctors. In round two, it won by 83% to 63%.
China’s medical AI strategy, released a year ago, called for its use in diagnosis, surgery, health monitoring through wearable devices and other applications.
In this nation’s overloaded medical system, AI is seen as a way of freeing pressured doctors from mundane tasks such as report writing, and increasing efficiency and diagnostic accuracy. Chinese hospitals analyze thousands of radiology images a day, a burden that increases the chances of misdiagnosis.
In 2015, China had 1.8 doctors per 1,000 people, compared with 2.6 in the U.S., according to data from the Organization for Economic Cooperation and Development. Austria has the highest proportion, 5.1.
According to Raymond Moh, chief executive officer of Hanalytics’ Beijing office, Biomind diagnoses diseases with 90% accuracy, without fatigue. He said sending machines to remote areas would be like deploying a top doctor. It could train inexperienced doctors by alerting them to conditions they might miss, he added.
“The role of AI is not to replace doctors but to help to investigate blind spots: ‘Please investigate further.’ For the more senior doctors, they see our AI machines as a way to speed up their work,” he said.
Medical studies in South Korea and elsewhere suggest that AI’s accuracy in diagnosing certain tumors is about the same as a senior doctor — but that doctors working with AI may produce better results.
Other AI programs are operating in more than 130 Chinese hospitals: Last year a robot developed by iFlytek and Tsinghua University passed China’s medical licensing exam with a score that was higher than 96% of candidates. Yitu Technology — a facial recognition specialist — is involved in a project in Chengdu, capital of Sichuan province in southern China, to diagnose cancer, while Chinese tech giants Alibaba, Tencent and Baidu are all involved in AI health projects.
Some of the competing doctors concluded Biomind would help them in their work.
“I would be more than happy to have one of these machines to help me,” said Fang.
Another competitor, Dr. Fang Weidong, said it would assist in the diagnosis of very early tumors that a doctor might miss.
“Every day I consult and diagnose over 100 patients, so you can really imagine the workload each doctor has every day. I’d welcome having AI to improve my efficiency,” he said.
Alexandra Golby, a professor of neurosurgery at Harvard, who was a judge in the competition, said AI was an important “second set of eyes” to assist doctors but could not replace them. It could add to their workload, rather than making things easier.
“I think the results of the challenge were very impressive. There’s no question about that. This is a very early stage and it’s moving incredibly quickly. It’s clearly a very powerful tool,” she said.
If doctors examined images after an AI machine, Golby added, there was a risk of confirmation bias, with doctors assuming the machine’s diagnosis was correct, overriding their own judgment.
Golby also raised the question about where medical liability would lie in case of a misdiagnosis — the doctor or the company that designed the machine?
The World Health Organization representative in China, Gauden Galea, said AI offered exciting possibilities, including the discovery of new syndromes and new treatments.
“The media like to scare about AI, reminding us of killer robots and stark futures dominated by machines in Hollywood movies,” he said. “I think we need to set those fears aside and focus on a real assessment of AI as a tool for clinical medical assessments.”