Artificial intelligence is changing the way we work, do school assignments, enjoy our airwaves and now … take sick days?
Anyone planning to fake an illness in the future might want to rethink their plans, as scientists are working on AI technology that will be able to detect whether a person is sick or not, based on the sound of their voice.
Researchers from Sardar Vallabhbhai National Institute of Technology in India have been studying how to diagnose illness using non-invasive techniques, and are in the process of developing a method that can identify when a person has a common cold, based on their speech sounds.
The study, as reported by The Economist and published earlier this year in the journal Biomedical Signal Processing and Control, is designed to help diagnose illness over the phone in an effort to minimize medical appointments and prevent the spread of viruses, but it might also make it harder for people to call in sick when they’re not actually ill.
The researchers, led by electronic engineer Pankaj Warule, analyzed the voice patterns of 630 German people, 111 of whom had a cold. They hypothesized that a cold could upset the vocal patterns normally present in those who are feeling well.
It’s no secret that voices change when a sore throat and runny nose come into play, and the researchers found that vocal rhythms of those with a cold are uniquely distinct, with differences in amplitude and frequency.
According to the study, subjects were asked to count from one to 40, before recounting what they did on the weekend. They were then asked to recite Aesop’s fable The North Wind and the Sun — a text frequently used when researching phonetics.
The new features proposed, which were placed into machine-learning algorithms, fared quite well in determining those who were feeling poorly; they were able to accurately classify cold and non-cold speech with scores of 69 per cent and 67 per cent, respectively.
As Insider points out, for the study’s results to be applied to the real world as-is, it would require a staff member to read the same fable while calling in sick, which sounds a little bit silly. Clearly, the research is still in the early stages and the methods and technology would need fine-tuning.
The study marks the latest in a long list of how AI is being used to remotely detect biomarkers for disease.
The Economist reports that doctors, computer scientists and psychiatrists are already researching how AI can help detect health problems based on how people walk, talk and write.
In a study published last year, researchers found that AI can be used to detect Parkinson’s disease by analyzing a person’s breathing patterns. And other studies have found that AI-trained analysis of vocal patterns can also detect depression and several forms of cancer.
In Canada, a lab out of Waterloo, Ont., has used AI to design an advanced form of magnetic resonance imaging (MRI) that could better help identify breast cancer markers and determine personalized treatment tailored to the individual and their type of cancer.
“The cancer itself just lights up and really shows the different nuances and characteristics around it, which makes it very much easier to identify not only where the cancer is, the size of the cancer, but also the actual tissue characteristics of the cancer to help doctors make better decisions,” Alexander Wong, professor and Canada Research Chair in Artificial Intelligence and Medical Imaging at the University of Waterloo, told Global News.