What medical experts found most interesting and innovative was the algorithm’s ability to access and assess data that could not be used earlier, like notes hidden in PDF files, and scribbles on old charts
Google’s new artificial intelligence algorithm can predict the probability of a patient dying more accurately than a doctor can, Bloomberg reported.
The conclusion was arrived at after woman with late-stage breast cancer checked into a hospital with fluid in her lungs and got a radiology scan done. The hospital computer analysed her vitals and predicted a 9.3 percent chance of her passing away during her stay there.
The same set of vitals was then run through Google’s new algorithm, which processed a total of 175,639 data points and predicted a 19.9 percent chance of the patient dying. Sure enough, the woman passed away after a few days.
The company had then published a harrowing account of the woman’s death and talked about the healthcare benefits of neural networks, a type of artificial intelligence software that analyses data to learn and evolve.
The company had created a tool that had the potential to forecast a host of patient-related outcomes, including how long they may spend in a hospital, the odds of them checking into the hospital again, and the probability of them dying.
What medical experts found most interesting and innovative was the algorithm’s ability to access and assess data that could not be used earlier, like notes hidden in PDF files, and scribbles on old charts.
Nigam Shah, who is associate professor at Stanford University, was quoted as saying that nearly 80 percent of the time spent today in predictive models is for the “scut work”, which involves making data presentable. However, Google’s software approaches the issue differently.
“You can throw in the kitchen sink and not have to worry about it,” Shah told Bloomberg.
