Artificial intelligence –
A tool assesses mental health and detects diseases by analyzing the voice
Application imagined at EPFL, Virtuosis is already used by companies concerned with the well-being of their employees. She could soon win the medical field.
The two founders of virtuosis, Edoardo Giudice and Lara Gervaise, with the interface of the application which assesses the level of stress, anxiety and happiness.
Florian Cella / Tamedia
- The virtuosis application analyzes the voice of its users to estimate their level of well-being.
- It could also serve as a screening tool for certain neurological, cardiometabolic and respiratory diseases.
- The founders explain that the tool is designed to support doctors, not to replace them.
- Like all projects that combine artificial intelligence and medical data, virtuosis asks ethical questions.
The principle of virtuosis is simple: you speak for about thirty seconds against your phone and the application tells you your level of stress, anxiety and happiness. It is the promise of this EPFL start-up, now based in the biopôle of unrest, which focuses on the analysis of the voice by artificial intelligence to do mental health prevention.
Launched two years ago, technology has attracted several companies that care about well-being of their employees. For them, the use is simple since it has been integrated into Teams, the Microsoft videoconference service. It is also available to the general public (app.virtuosis.ai) with an interface that offers a regular check-up. Finally, high -level athletes are also starting to be interested in it.
The founders of virtuosis, Lara Gervaise and Edoardo Giudice, see further: “So far, we have focused on non -medical aspects. Virtuosis makes it possible to assess well-being, to prevent certain psychic difficulties and to work on your communication. Now we also make it a screening tool And support for diagnosis for certain pathologies, in particular neurological, cardiometabolic and respiratory diseases ”.
How does it work?
When he deciphers the spectrum of the voice, the system reviews several dozen indicators, “vocal biomarkers” such as tone, rhythm or monotony. “The combination of these parameters provides the user’s psychic state or suggests certain pathologies,” explains Lara Gervaise. We know for example that diabetes can impact the nervous system at the jaw and promote acid reflux which can damage vocal cords. This induces changes in their voice and they are detectable. ”
According to the founders, the scope is vast since virtuosis could identify the warning signs of around thirty psychic or somatic diseases. “We get very good results for the detection of Parkinson’s disease. It is possible to identify the tremors in the voice even when they are tiny and not perceptible in the human ear. ”
“In the case of certain neurological diseases, it can be quite easy to detect a mechanical problem that translates into the voice,” confirms Professor Idris Guessous, head of the Hug innovation center. For mental disorders, it is more complicated, you have to analyze the way the person engages in the discussion. In case of depression, the voice will generally be flatteer, monotonous. ”
Another example of use: detection of menopause. “Hormonal changes induce changes to the larynx. Science has identified it for a long time. What changes with virtuosis is that the deep learning (Editor’s note: process that allows the machine to learn from large amounts of data in order to solve complex problems) improves the reliability of detection, ”reports Lara Gervaise.
Still limited to mental health
If the tool is reliable for many pathologies, why stick to mental health and well-being in the consumer version? “On the one hand, we continue to collect clinical data in order to consolidate the system. On the other hand, in terms of public health, you have to ask yourself whether it is relevant to put such a tool in everyone’s hands, ”replies Edoardo Giudice.
“The goal is not to short-circuit the medical system, but to integrate it. Receiving a diagnosis on your screen, without being accompanied and without knowing the possibilities of care, is not desirable, ”completes his partner.
Professor Idris Guessous is responsible for the Hug innovation center. “If the machine becomes able to recognize emotions by analyzing a voice, it is a very great advance,” he says.
DR
“The arrival of AI in the medical world is a good thing, provided that it is not there to replace the doctor, but to accompany him,” says Professor Idris Guessous. With this software, we see that technology can invest in the areas that we thought reserved for humans. So far, the computer was very strong to do things that are complicated for us, for example, quickly summarizing a 100 -page scientific relationship. On the other hand, he struggled with things that seem simple to us like detecting emotions, which a child can already do. If the machine becomes capable of recognizing emotions by analyzing a voice, it is a very great advance. ”
Open ethical questions
Like all projects that combine artificial intelligence And medical data, virtuosis asks ethical questions. The officials say they also want to stand out on this plan. During a videoconferencing, for example, they emphasize that the system only analyzes the voice of the one who activates it, without dwelling on what is said and without worrying about the image.
This does not necessarily prevent malicious uses. Why not imagine a company which, during a job interview, uses it to detect possible psychic vulnerabilities among candidates? “From the start, we have received requests that do not agree with our values or for uses that do not please us. We systematically refuse them, says Edoardo Giudice. This is also why all the technology is not free access. We want to collaborate with institutions that will make positive use of it and we think a lot about how to limit abuses. We are also open to partnerships with foundations or non -profit organizations. ”
Ultimately, Virtuosis hopes to contribute to the strengthening of screening strategies and improving access to care, especially in regions where doctors are few. “In some countries, medical deserts lead to a significant appeal to telemedicine. It is not a bad thing, but a disease like Parkinson needs a face -to -face diagnosis. A tool that detects the beginnings and gives rise to early management can change many things for patients ”, boasts Lara Gervaise.
Did you find an error? Please report it to us.