AI predicts patients' lifespans as well as a doctor
At some point we've all succumbed to the lure of a terrible online quiz that purportedly tells you how long you will live after asking a set of inane questions. But imagine if a seriously intelligent artificial intelligence was asking the questions, and it could accurately predict how long you have left to live? A team at the University of Adelaide in Australia has been working to create an AI system that can predict a person's lifespan just by studying images of their organs, and its early results show its predictions are just as good as a human doctor.
In a first of its kind study, using artificial intelligence to determine lifespan from medical images, the research used AI to examine the medical images of 48 patients' chests.
The AI then predicted which patients would die within five years, with an accuracy rate of 69 percent. This rate is comparable to similar "manual" predictions by doctors.
While a study like this has potentially frightening implications, with hyperbolic prognosticators easily imagining a utilitarian future where "high death risk" patients could be denied treatment, the researchers offer more positive directions for the technology.
"Predicting the future of a patient is useful because it may enable doctors to tailor treatments to the individual," says lead author Dr Luke Oakden-Rayner. "Instead of focusing on diagnosing diseases, the automated systems can predict medical outcomes in a way that doctors are not trained to do, by incorporating large volumes of data and detecting subtle patterns."
This study follows on from a Chinese startup recently revealing it had developed an AI system that could assist in examining CT scans diagnosing lung cancer. In addition to that, we saw IBM's AI system Watson enter hospitals last year and help answer patient questions. It's clear that intelligent machines are speedily moving into our medical system.
The University of Adelaide team is set to expand the AI's focus, moving its analysis to tens of thousands more patient images. If at this early stage the machine is accurately predicting patient's lifespans at a rate similar to a human doctor, we can only imagine what future, more learned iterations of the system could do.
The new research was published in the journal Scientific Reports.
Source: The University of Adelaide