In a groundbreaking advance at the crossroads of neuroscience, artificial intelligence, and digital health, researchers at the University of Geneva (UNIGE) have demonstrated that everyday smartphones and smartwatches can serve as powerful tools for the early detection of neurological and mental health disorders. By integrating continuous passive data collection through wearable devices with sophisticated AI algorithms, the team has unveiled a non-invasive, scalable method capable of predicting emotional and cognitive fluctuations with remarkable accuracy. This pioneering study, recently published in npj Digital Medicine, could redefine preventative brain health strategies and herald a new era where technology proactively monitors the earliest signs of brain dysfunction.
Brain health is an urgent global priority, as neurological disorders and mental illnesses afflict vast populations worldwide. The World Health Organization reports staggering statistics: over one in three individuals experience neurological disorders such as stroke, epilepsy, or Parkinson’s disease, while more than half will endure mental health challenges, including depression, anxiety, or schizophrenia, at some stage in their lives. These figures are projected to escalate significantly with increasing global longevity. Traditional clinical approaches, reliant on episodic testing and self-reporting, often miss subtle, early changes in cognitive and emotional functioning — changes that, if detected sooner, could inform timely interventions and mitigate disease progression.
Responding to this challenge, the UNIGE team embarked on an ambitious longitudinal study aimed at discovering if ubiquitous connected devices could continuously and objectively assess brain health markers over extended periods in real-world settings. The study enlisted 88 volunteers aged 45 to 77, a demographic representative of middle-aged to older adults, who are at increasing risk for neurodegenerative and psychiatric conditions. Each participant was equipped with a smartwatch and a bespoke smartphone application designed to harvest “passive” data streams without altering their daily routines. These passive signals encompassed physiological metrics such as heart rate, physical activity patterns, sleep behavior, and external environmental factors including weather conditions and air pollution levels — resulting in 21 diverse indicators being collected continuously across ten months.
Complementing these passive measurements, participants provided “active” data every three months through self-administered questionnaires targeting emotional well-being and neurocognitive tests assessing various aspects of cognition. This dual-modal approach created a rich dataset capturing both objective biological-environmental variables and subjective psychological-cognitive states. The core innovation lay in employing advanced AI to analyze this multifaceted data, aiming to correlate subtle patterns in the passive signals with the more explicit mental health and cognitive state assessments.
Igor Matias, a doctoral assistant at UNIGE’s Research Institute for Statistics and Information Science and the study’s lead author, explained the central hypothesis: Could artificial intelligence discern meaningful predictors of brain health by mining continuous data from wearable devices? The results surpassed expectations. AI models achieved an average prediction error rate of just 12.5% when forecasting fluctuations in participants’ cognitive and emotional states, marking a significant leap toward reliable early detection of brain health deviations without intrusive clinical procedures.
Delving deeper, the study revealed that AI predicted emotional states with the highest precision, registering error rates generally between 5% and 10%. This suggests that passive physiological and environmental markers resonate strongly with transient emotional conditions and affective states. Cognitive states, inherently more complex and multifactorial, were predicted with slightly less accuracy, yielding error rates from 10% to 20%. These findings underscore the relative ease with which AI can decode emotional health signals, perhaps due to more direct links between autonomic physiological responses and emotional stimuli compared to the more nuanced and composite nature of cognitive processes.
Crucially, the researchers identified key passive indicators that were most informative in driving the AI predictions. For cognition, environmental air pollution levels, weather conditions, daily heart rate variability, and sleep quality metrics stood out as influential. In contrast, emotional state predictions heavily relied upon weather parameters, sleep variability indices, and heart rate measurements during sleep cycles. These insights not only validate previous research highlighting the profound impacts of external environment and lifestyle on brain function but also provide tangible biomarkers for continuous monitoring.
The study’s integration of environmental data alongside physiological parameters represents a pioneering holistic approach, recognizing that brain health is dynamically influenced by an interplay of intrinsic biological factors and extrinsic environmental exposures. This multidimensional perspective enhances predictive fidelity and paves the way for contextualized, personalized brain health management in everyday life. Wearable sensors thus transcend mere convenience gadgets, evolving into vital digital biomarkers capable of mapping neurological and psychological health trajectories at unprecedented temporal resolutions.
Supervised by Professors Katarzyna Wac and Matthias Kliegel, experts in digital health and cognitive aging respectively, this study forms part of the wider Providemus alz project, which aims to harness mobile technology for neurodegenerative disease monitoring and management. Encouraged by the promising results, the team is now embarking on an extended 24-month data collection phase. This next stage will delve into individual variability linked to AI model performance, exploring why certain participants’ data yield stronger predictive insights than others. Through this, the research objectives extend toward refining AI algorithms to accommodate personal traits and lifestyle nuances, ultimately enabling tailored preventative strategies and interventions.
The implications of this research are profound and manifold. By deploying AI-driven continuous brain health monitoring through everyday wearable devices, the boundaries between clinic and daily life begin to blur. Early detection of subtle cognitive decline or emotional dysregulation becomes feasible without the need for specialized infrastructure or frequent clinical visits. This democratization of brain health surveillance promises enhanced patient empowerment, more timely therapeutic responses, and potentially reduced burdens on healthcare systems increasingly strained by neuropsychiatric disorders.
At a technical level, the study highlights how machine learning models can effectively synthesize heterogeneous streams of time-series data — physiological, environmental, behavioral — spanning multiple temporal scales. Key challenges addressed include noise reduction, missing data imputation, temporally aligned feature extraction, and algorithmic interpretability. The successful integration of these complex data modalities into actionable predictions marks a significant methodological achievement, setting a benchmark for future digital biomarker research.
Moreover, this work underscores the importance of passive data collection paradigms that do not impose behavioral changes or active compliance burdens on users. By unobtrusively capturing continuous signals, the approach maximizes ecological validity and participant adherence, critical factors for long-term monitoring endeavors that seek real-world applicability beyond controlled experimental settings.
In conclusion, the University of Geneva’s study heralds a new frontier in brain health assessment, showcasing how everyday smart devices coupled with AI can detect early signs of neurological and mental illnesses with remarkable precision. As the global burden of brain disorders surges, such innovations offer hope for scalable, proactive, and personalized mental healthcare solutions. The scientific community and technology developers alike will watch keenly as this promising approach advances toward broader clinical validation and eventual integration into routine health monitoring protocols.
Subject of Research: People
Article Title: Digital biomarkers for brain health: passive and continuous assessment from wearable sensors
News Publication Date: 3-Mar-2026
Web References: http://dx.doi.org/10.1038/s41746-026-02340-y
Keywords: Digital biomarkers, brain health, artificial intelligence, wearable sensors, smartphones, smartwatches, cognitive health, emotional health, passive data collection, machine learning, neurodegenerative diseases, mental illness detection
