AI can predict if you'll die soon – but we've no idea how it works
AI can predict if you'll die soon – but we've no idea how it works
Artificial intelligence can
predict a person’s chances of dying within a year by looking at heart test
results – even when they look normal to doctors. How it does so is a mystery.
Brandon Fornwalt at healthcare provider
Geisinger in Pennsylvania, US and colleagues tasked an AI with examining 1.77
million electrocardiogram (ECG) results from nearly 400,000 people to predict
who was at a higher risk of dying within the next year.
An ECG records the electrical activity of the
heart. Its pattern changes in cardiac conditions including heart attacks and
atrial fibrillation.
The team trained two versions of the AI: in one,
the algorithm was only given the raw ECG data, which measures voltage over
time. In the other, it was fed ECG data in combination with patient age and
sex.
They measured the AI’s performance using a
metric known as AUC, which measures how well a model distinguishes between two
groups of people – in this case, patients who died within a year and those who
survived. The AI consistently scored above 0.85, where a perfect score is 1 and
a score of 0.5 indicates no distinction between the two groups.
Read more: Mind
meld – Artificial intelligence is improving the way humans think
The AUCs for risk scoring models currently used
by doctors range between 0.65 and 0.8, says Fornwalt.
For comparison, the researchers also created an
algorithm based on ECG features that doctors currently measure, such as certain patterns from the recordings.
“No matter what, the voltage-based model was
always better than any model you could build out of things that we already
measure from an ECG,” says Fornwalt.
The AI accurately predicted risk of death even
in people deemed by cardiologists to have a normal ECG. Three cardiologists who
separately reviewed normal-looking ECGs weren’t able to pick up the risk
patterns that the AI detected.
“That finding suggests that the model is seeing
things that humans probably can’t see, or at least that we just ignore and
think are normal,” says Fornwalt.
Read more: How
AI is ushering in a new era of superhuman doctors
“AI can potentially teach us things that we’ve
been maybe misinterpreting for decades,” he says.
It’s still unclear what patterns the AI is
picking up, which makes some physicians reluctant to use such algorithms.
This research was based on historical data, and
it will be important to demonstrate in clinical studies that such an algorithm
improves patient outcomes, says collaborator Christopher Haggerty.
The research will be presented at the American
Heart Association’s Scientific Sessions in Dallas on November 16.
Comments
Post a Comment