Polish experts have developed a system that detects depression, anxiety and other disorders based on eye movements in just 10 seconds. The accuracy currently reaches 70 percent, but researchers see the possibility of increasing it further.
Scientists from three Polish universities have created am AI-based system for the rapid detection of mental disorders.
A total of 101 people participated in the study leading to the development of the system, among them - patients diagnosed with depression, people with social anxiety and healthy people constituting the control group. The task of the subjects was to look at photos of faces showing different emotions for 10 seconds. Special sensors placed in eye trackers recorded the movements of the volunteers' eyes. The collected data were used to generate the gaze paths, which were then analysed by neural networks.
'Eye movement patterns can provide objective data on our mental condition. In people with depression, we notice a tendency to focus on negative stimuli. In turn, people with social anxiety show increased activity in facial scanning, which is related to the occurrence of a phenomenon called hyperscanning. This is manifested by an extension of the facial scanning path, which indicates the subject's sensitivity to social stimuli', says Karol Chlasta, PhD, from the Leon Kozminski Academy, co-author of the study and expert in the field of artificial intelligence.
Psychologists and other AI research experts were also involved in creating the system, including Krzysztof Krejtz, PhD, and Izabela Krejtz, PhD, from SWPS University, and Katarzyna Wisiecka, PhD, from the VIZJA University in Warsaw.
In the case of depression and social anxiety, the method's effectiveness is 60-70 percent. The researchers emphasise that this value is comparable to traditional methods.
At the same time, the newly developed approach is much faster and less engaging for the patient than traditional methods. It also enables easier and faster monitoring of changes in a person's mental state, the authors of the solution emphasise.
The new method can be implemented in everyday devices - for example, a laptop, smartphone or VR goggles. 'We can compare it to the operation of smart watches, which tell us what our sleep rhythm was like. Except that in addition to the pulse, we can also measure and analyse our eyesight, for example', Chlasta explains.
This approach is expected to enable the creation of a 'mental well-being graph', similar to the graphs showing steps, heart rate or sleep quality offered by today's smart watches and smartphones.
Scientists do stop at vision analysis. They are also working on the application of new artificial intelligence methods to analyse voice - in the context of diagnosing both depression and neurological disorders.
'Many diseases slightly change our voice. It can be compared to a computer that is overloaded and works a bit slower. It has to switch between different tasks. In the case of people, we can observe changes in the functioning of the speech organs, which are often difficult for humans to detect, but which a system based on artificial neural networks is able to immediately pick out, even from short fragments of speech', Chlasta says.
'This can give an early warning that you are at risk of depression, dementia or Alzheimer's disease. Thanks to this, you have a chance to react faster. Change your daily habits and consult a doctor before your symptoms get worse', the scientist adds.
The creators of the system emphasize that depression and social anxiety are among the most common mental disorders, and the number of people affected by them continues to grow. According to WHO forecasts, by 2030, depression will be the most frequently diagnosed disease in the world. In Poland, approx. 4 million people already suffer from it, although many cases remain unnoticed. One of the main reasons for the lack of diagnoses is that early detection of depression is difficult and time-consuming.
A quick analysis of eye movements, however, can provide valuable information about the mental state and be an important signal to consider consulting a doctor, the authors of the solution suggest. They also point out that further research will be necessary for the wide introduction of such a new method. 'Additional data are needed to create a well-functioning model. However, they are not collected systematically, and medical facilities do not collect them either. Additionally, public trust in artificial intelligence is still low, and people approach the provision of their data with a certain amount of uncertainty. The truth is, however, that without additional data it will not be easy for us to go beyond laboratory conditions, and without further analyses we will not be able to demonstrate the prototype of this new technology in operational conditions', Chlasta explains.
System-wide changes are also needed. According to Chlasta, it would be possible to use the technology developed by his team in a broader system that would enable monitoring of the mental health of patients - be it on a scale of the entire healthcare system or, for example, at the level of a single clinic.
Find out more on the website. (PAP)
Marek Matacz
mat/ zan/
