Doctors commonly use "risk scores" to help them decide on treatment. Computers are better at calculating complicated risk factors
All the days ordained for me were written in your book before one of them came to be. -- Psalm 139:16
(dpa) – It's a brave new world when computers can predict our death, but such a world might not be out of the realm of reality, according to European researchers.
A study presented at an international cardiology conference in Lisbon this month found that artificial intelligence can already do a better job than humans of calculating a person's risk of death.
And not just a slightly better job, either. A computer algorithm managed to identify patterns in patients that had died with 90-per-cent accuracy, the researchers contend.
Judging the risk of death is a common if morbid practice for doctors, who will use "risk scores" to help them decide on treatment. The point of the study was to show just how many more risk factors a computer can identify and use to develop such a score than humans.
The research used machine learning - the backbone of AI that essentially involves submitting reams of data to a computer that can learn how to piece it all together – to predict outcomes around heart disease.
The study took 950 patients with chest pains and sent them for a typical round of tests to see if they had coronary artery diseases, identifying 85 different variables based on those tests and their general medical records. The study then followed up six years later, after which there had been 24 heart attacks and 49 other deaths in the group.
"By repeatedly analyzing 85 variables in 950 patients with known six-year outcomes, an algorithm 'learned' how imaging data interacts. It then identified patterns correlating the variables to death and heart attack with more than 90 per cent accuracy," according to a statement on the study.
Researchers said the findings could help doctors improve treatments by better identifying the risks of death in patients, though they admitted the findings carried their own risks if misused and would still take some time to evaluate.
"These advances are far beyond what has been done in medicine, where we need to be cautious about how we evaluate risk and outcomes," said Luis Eduardo Juarez-Orozco of the Turku PET Centre in Finland and the study's author.
"We have the data but we are not using it to its full potential yet," Juarez-Orozco said. Some science-fiction fans might say that's a good thing.