午夜福利1000集合

Time to start hoping a robot takes that emergency call. When it comes to spotting signs of cardiac arrest, artificial intelligence is beating humans.

New Scientist first covered Copenhagen-based artificial intelligence company Corti earlier this year.聽Its AI was being trialled in Denmark to listen in on emergency calls in real-time. It searches for patterns of communication, including features like tone of voice and breathing sounds. Now we know how well the system works.

According to a study of 161,650 emergency calls using Corti鈥檚 system, it is more accurate and faster at detecting signs of a cardiac arrest over the phone than dispatchers.

The AI correctly detected 93 per cent of cardiac arrests compared to the dispatchers鈥� detection rate of 73 per cent. It took only 48 seconds on average to detect the condition, compared to dispatchers who took 80 seconds.

鈥淓very second counts when recognising cardiac arrest and we know that reducing downtime increases chance of spontaneous circulation, morbidity post arrest and survival, so studies into improving out-of-hospital cardiac arrest outcomes can only be a positive thing,鈥� says Natalie Cookson, an emergency medical trainee doctor, working in London.

However, she adds the study focusses on improving time to recognition of cardiac arrest and delivery of resuscitation, which may only improve factors such as hypoxia and cardiac output. Other complications associated with cardiac arrest are not addressed in the study, she says, so 鈥渇urther studies will be needed to evaluate actual patient outcomes鈥�.

Working with medics

鈥淭he technology works so well on out-of-hospital cardiac arrest, since this is a very binary condition which usually presents itself in a very specific way,鈥� says Stig Nikolaj Blomberg at the University of Copenhagen, who lead the study.

Instead of replacing emergency dispatchers, the system is designed to be used alongside them. 鈥淚t all comes down to the idea of supporting humans with machine learning, and supplementing them to do even more,鈥� says Andreas Cleve, Corti鈥檚 CEO.

The algorithm is already being used in Copenhagen, and the company has announced a new partnership with Belgian non-governmental organisation the European Emergency Number Association (EENA) to spread it across the continent. Four sites across Europe will be selected to pilot the technology for six months, and the results of this will be shared in 2019.

Meanwhile, the research team in Copenhagen is now planning to test the technology聽on detecting signs of a stroke.