AI-backed system prevents dozens from killing selves

By Xu Hailin Source:Global Times Published: 2018/11/28 22:28:40

A medical system using artificial intelligence (AI) technology has successfully prevented more than 40 people from dying by suicide since it came online in July, according to a renowned scientist at a recently concluded conference in Boao, South China's Hainan Province.

Huang Zhisheng, chief scientist of the AI team for cerebral protection at the Beijing-based Capital Medical University, started a medical care project called Shudong Xingdong, or Tree Hole Operation, using AI to detect people who are at the brink of killing themselves, the Science and Technology Daily reported on Wednesday.

The system has an average accuracy rate of 82 percent after 99 percent of irrelevant information is eliminated during its analysis, the report said.

More than 300 people considered suicide high risks have been followed by the system as of press time.

A medical assistance team, consisting of more than 100 volunteers and mental health experts, has reached more than 40 people, prevented them from killing themselves and helped them regain hope in life, according to the report.

"Similar attempts have been made before AI technology was adopted. But due to limited computing power, the technology couldn't be applied," Yang Bowen, an algorithm engineer at Yitu Technology, a Shanghai-based company that focuses on AI research and innovation, told the Global Times on Wednesday.

Thanks to the computing power of AI technology, the concept was realized, Yang said.

The AI system would play a more important role in preventing suicides and treating patients suffering from depression, Wang Gang, head of the Beijing Anding Hospital and member of the operation's medical team, was quoted as saying by Science and Technology Daily.

About 60 million people in China are suffering from depression.

Those in the most serious stage of the mental disease are highly prone to killing themselves, the report said.

Posted in: SOCIETY

blog comments powered by Disqus