This research, published in Pervasive and Mobile Computing, was conducted with data from adult participants in Pakistan, whereby movements were recorded with the use of a sensor as individuals performed a series of activities in a particular order. These strides with technology have the potential to open up new possibilities for addressing mental health challenges more effectively.
Detecting Anxiety Through Behavior
This study was conducted with ten participants, between the ages of 20 and 50 years, with motion sensors used to detect behavioral signs of anxiety. Researchers focused on specific behaviors to detect anxiety using motion sensors and deep learning techniques, including nail-biting, knuckle cracking, hand tapping, etc, which were found to be over 92% effective. While this research provides insights into how AI can be used to better assess for symptoms of anxiety, a limitation of this study is its small data set, as it was only conducted with a total of 10 participants.
AI Offers Mental Health Possibilities
Social psychologist, and researcher for this study, Gulnaz Anjum, PhD, says, “A major takeaway from this research is that we can safely and conveniently use artificial intelligence (AI) to provide measurement, analysis, and diagnostics for anxiety.” Another of the study’s researchers, Nida Saddaf Khan, MS, MBA, explains, “For human activity recognition, deep learning is among the most reliable and robust algorithms. It has earned the researchers’ trust due to its capability to learn the temporal dynamics and complex patterns even from raw sensors’ data.” Anjum notes that this can be as convenient as wearing a smartwatch with their app and checking the readings, once it becomes available. “We are working with a higher number of human participants to establish comparisons but the initial research provides a safe, non-intrusive, non-subjective, and accurate way of measuring anxiety,” she says. When measuring anxiety using only subjective measures, scales, and assessments, Anjum highlights how they can cause further stress and anxiety among both subjects and clinicians. “Using AI, such as our sensors and deep learning models can be very helpful because this assessment works in the background without our conscious attention,” she says. Anjum explains that people should remember while reading and applying their research that the goal is to expand horizons for the identification of anxiety disorders and ultimately, the improvement of people’s mental health, but cautions that their work is only relevant for diagnostics. In this way, Anjum recommends connecting with a mental health practitioner for any post-diagnostic work and support. “AI is a brilliant tool and a reliable indicator for identification of our anxiety, but when it comes to seeking help, we really need to reach out to a clinician,” she says. Anjum notes that measurement of any behavioral aspects, if done correctly and in collaboration with field experts of mental health is possible now. “We have brought forward the first examples of this kind and we are sure that the role of AI in psychological assessment is the future of safe and accessible mental health for all,” she says. Especially during the pandemic, Anjum says, “Many more people around us have been experiencing higher levels of anxiety due to COVID-19 and climate shocks around the world, so the need for easier and nonintrusive modes of measuring anxiety is higher than ever before.”
A Promising Mental Health Tool
A psychiatrist with Mindpath Health, Rashmi Parmar, MD, says, “This is a unique study in which the authors utilized human activity recognition sensors paired with smartphones to detect certain physical movements which are associated with anxiety in humans.” Dr. Parmar explains that the eventual goal is to provide more accurate data for clinical researchers and doctors to identify, evaluate and treat anxiety disorders. “Anxiety can manifest in different forms in different people which could include physical as well as emotional symptoms,” she says. While activities measured in this study may reflect anxiety, Dr. Parmar notes they may not always be an outcome of anxiety. “For example, other scenarios like fatigue or boredom can also lead to similar activities and may be indistinguishable from anxiety unless clinically evaluated,” she says. Dr. Parmar explains, “This study is a useful starting point. The findings of this study need to be correlated clinically for more accuracy. Although AI can help identify at-risk individuals, a detailed clinical evaluation will still be required to confirm a psychiatric diagnosis.” Since AI research in mental health is still in its infancy, Dr. Parmar highlights that there are not many studies for comparison. “When you consider AI and its application in medicine, it is important to balance the overall risks vs. benefits of the designed application and whether the results can be replicated in the real world,” she says. Dr. Parmar notes, “AI sounds like a promising tool in the future of mental health, especially with electronics and smart phones being so popular with the current generation. If designed well, AI tools can aid in early detection, evaluation, and treatment of psychiatric illnesses and may possibly help with prevention efforts as well.” From personal experience, Dr. Parmar explains that it may be possible to identify anxious individuals based on certain verbal and non-verbal cues, as a person may have a tense posture, trembling, a nervous handshake or smile, increased sweating, rapid breathing, difficulty thinking, or maintaining a steady conversation. Dr. Parmar explains that a diagnosis can be confirmed by a thorough clinical interview and mental status exam. “If designed properly, AI-based tools can come in quite handy in everyday clinical scenarios,” she says. While AI will not be able to replace the much-needed “human touch” in medicine, Dr. Parmar notes that it definitely can make life easier and reduce burnout for those dedicated to shaping a better future for medicine.
