Smartwatches capable of collecting physical and physiological data may serve as valuable instruments in biomedicine, particularly for comprehending brain diseases and behavioral disorders, as well as identifying potential driver mutations linked to these conditions. This assertion is made in a study published in the journal Cell, co-authored by Mark Gerstein from Yale University and involving Professor Diego Garrido Martín from the University of Barcelona.
The research utilized smartwatch data from over 5,000 adolescents to train artificial intelligence models aimed at predicting various psychiatric illnesses and identifying associated genes. The findings indicate that these wearable sensors could facilitate a more nuanced understanding and treatment of psychiatric conditions.
Mark Gerstein, an expert in biochemistry, computer science, statistics, and data science, stated, "In traditional psychiatry, a doctor will assess your symptoms and you'll either be diagnosed with an illness or won't. But in this study, we focused on processing the wearable data in a way that could both be leveraged to predict illnesses more comprehensively, and to better connect them to underlying genetic factors."
Quantitative illness detection poses challenges; however, wearable sensors that collect continuous data may provide solutions. The study leveraged data from the Adolescent Brain Cognitive Development Study, the largest long-term assessment of brain development and child health in the United States. Data collected from smartwatches worn by adolescents aged 9-14 included heart rate, calorie expenditure, physical activity intensity, step count, sleep level, and sleep intensity.
Jason Liu, a researcher in Gerstein's lab and co-lead author of the study, remarked, "When processed correctly, smartwatch data can be used as a 'digital phenotype.'" The term 'digital phenotype' refers to traits measurable and trackable through digital tools like smartwatches.
Liu further explained, "One advantage of doing this is that we can use the digital phenotype almost as a diagnostic tool or a biomarker, and also bridge the gap between disease and genetics."
The research team developed a methodology for acquiring and converting the vast amount of smartwatch data into usable information for AI model training, which Gerstein described as a "new problem to solve in the research world which is technically challenging."
The study revealed that heart rate served as the most significant predictor for ADHD, while sleep quality and stage were more crucial for identifying anxiety. Gerstein noted, "These findings suggest that smartwatch data can provide us with information about how physical and behavioral temporal patterns relate to different psychiatric illnesses."
Additionally, the data may assist in differentiating various subtypes of psychiatric disorders. Beatrice Borsari, a postdoctoral associate at Gerstein's lab, stated, "For example, within ADHD there are different forms. Maybe we can extend this work to help distinguish between forms of inattention and hyperactivity, which typically respond to different pharmacological treatments."
Having established that the digital phenotype could predict psychiatric illnesses, the team further explored its potential in identifying underlying genetic factors using multivariate statistical tools developed with contributions from the University of Barcelona.