{"slug": "usc-study-used-sweat-brain-signals-eye-movements-and-ai-to-detect-depression-and", "title": "USC Study Used Sweat, Brain Signals, Eye Movements and AI to Detect Depression and Suicide Risk", "summary": "USC researchers developed an AI-powered assessment that analyzes biomarkers including sweat, brain signals, and eye movements to classify depression and suicidal ideation. The DARPA-funded PRECOG project aims to provide psychiatrists with objective tools to identify high-risk individuals and enable timely intervention, addressing the lack of standard testing methods for mental health conditions.", "body_md": "Suicide rates in the U.S. have been rising at an alarming rate over the past few decades, with rates among veterans being 1.5 times higher than those of the general public, [USC Suzanne Dworak-Peck School of Social Work](https://dworakpeck.usc.edu/)‘s [study](https://dworakpeck.usc.edu/news/upstream-suicide-prevention-research-demonstrates-importance-of-looking-beyond-mental-health) shows.\n\nWhile research shows that intervention is key to preventing suicide, many cases go unreported, particularly within military populations, due to societal pressure, stigmas around mental health and biases in reporting.\n\nA major root cause behind these challenges is that mental illnesses like depression still lack a standard testing method, as diagnoses today still mainly rely on self-reporting tools such as surveys and clinical interviews. This remains a critical challenge across mental illnesses.\n\nResearchers and clinical psychiatrists face an urgent need to develop objective testing metrics and tools that can help classify mental health conditions, as well as predict suicidal behaviors.\n\nIn a recent study, USC researchers with backgrounds across engineering, neurology, speech technology, artificial intelligence (AI), linguistics and clinical psychiatry teamed up to develop an assessment that classifies depression and suicidal ideation by analyzing biomarkers such as sweat, brain signals and eye movements, and using AI to build this classification system. The study was led by [Shrikanth Narayanan](https://viterbi.usc.edu/directory/faculty/Narayanan/Shrikanth), USC’s inaugural vice president for presidential initiatives and University Professor, with joint appointments across [USC Viterbi School of Engineering](https://viterbischool.usc.edu/)’s [Ming Hsieh Department of Electrical and Computer Engineering](https://minghsiehece.usc.edu/) and [Thomas Lord Department of Computer Science](https://www.cs.usc.edu/), as well as the [USC Mark and Mary Stevens School of Computing and Artificial Intelligence](https://stevens-computing-ai.usc.edu/).\n\nThis research project could provide psychiatrists with a tool to identify and potentially predict high-risk individuals through biological signals, offering a more standardized way to monitor mental health and identify timely intervention points to help prevent suicide.\n\nFunded by the [Defense Advanced Research Projects Agency (DARPA)](https://www.darpa.mil/), the research project is titled “[PRECOG: Multimodal integration of neural and biobehavioral signals for predicting preconscious responses](https://sail.usc.edu/~precog/)” and [began in June 2023](https://viterbischool.usc.edu/news/2023/07/usc-to-lead-interdisciplinary-project-on-mental-health/).\n\nThe PRECOG project has resulted in five papers published in major journals this year, with one additional paper currently in the publication process, covering additional biomarkers and computational breakthroughs.\n\nOne paper focused on tracking eye movements, and was published in npj Digital Medicine. Titled “[Deep Learning Characterizes Depression and Suicidal Ideation in Young Adults From Eye Movements](https://www.nature.com/articles/s41746-026-02550-4),” this paper was led by [Kleanthis Avramidis](https://klean2050.github.io/), a PhD student in the [USC Signal Analysis and Interpretation Laboratory (SAIL)](https://sail.usc.edu/), and advised by Narayanan.\n\nThree papers focused on neural signals measured using electroencephalography (EEG). One, titled “[Neural Evidence of Disrupted Self-Referential Processing in Suicidal Depression](https://www.sciencedirect.com/science/article/abs/pii/S0165032726006683),” was published in the Journal of Affective Disorders and led by PhD student [Colin McDaniel](https://dornsife.usc.edu/psyc/profile/colin-mcdaniel/), who is advised by Narayanan and [Assal Habibi](https://dornsife.usc.edu/profile/assal-habibi/), who is an associate professor of psychology at [USC Dornsife College of Letters, Arts and Sciences](https://dornsife.usc.edu/). Another, titled “[Time-Resolved EEG Decoding Reveals Altered Neural Dynamics of Affective Semantic Evaluation in Depression and Suicidality](https://www.nature.com/articles/s42003-026-10108-z),” was published in Communications Biology and led by PhD student [Woojae Jeong](http://linkedin.com/in/woojae-jeong-607822267), who is part of [USC Biomedical Imaging Group (BIG)](https://neuroimage.usc.edu/neuro/home) and advised by [Richard Leahy](https://viterbi.usc.edu/directory/faculty/Leahy/Richard). Leahy is the USC Leonard Silverman Chair and professor of electrical and computer engineering, biomedical engineering, and radiology. The third paper, titled “[Neural Responses to Affective Sentences Reveal Signatures of Depression](https://arxiv.org/abs/2506.06244),” was published in Translational Psychiatry and led by PhD student [Aditya Kommineni](https://www.linkedin.com/in/kommineniaditya/), who is also advised by Narayanan.\n\nThe fifth paper, titled “[A Point Process Model of Skin Conductance Responses in a Stroop Task for Predicting Depression and Suicidal Ideation](https://arxiv.org/abs/2510.00422),” used electrodermal activity (EDA), known as skin conductance, to collect sweat data as a biomarker. The study was also led by Avramidis.\n\nThe study was made possible by USC’s multidisciplinary collaboration across fields, with linguistics faculty [Dani Byrd](https://dornsife.usc.edu/profile/dani-byrd/) and clinical psychiatry faculty [Rael Cahn](https://dornsife.usc.edu/cmbs/rael-cahn/) also serving as co-authors on the papers. Byrd is a a professor at [USC Dornsife](https://dornsife.usc.edu/)‘s[ department of linguistics](https://dornsife.usc.edu/ling/). Cahn is a clinical associate professor of psychiatry and the behavioral sciences, with a joint appointment between [USC Dornsife](https://dornsife.usc.edu/) and [Keck School of Medicine of USC](https://keck.usc.edu/).\n\n## Measuring Depression with Biological Signs From Our Skin, Eyes and Brains\n\nThe PRECOG study aimed to move mental health diagnosis toward evidence-based medicine, where clinicians can base decisions on physiological data rather than conversations and self-reported symptoms alone.\n\nAcross the currently published papers, researchers examined multiple biomarkers, including neurological signals recorded through EEG, EDA and behavioral indicators captured through eye-tracking. Using the collected data, engineers on the team applied AI and deep learning techniques to build classification models capable of identifying stable markers of depression.\n\nThese classifications were developed by analyzing how the skin, brain and eyes respond to specific emotional stimuli, allowing researchers to measure how an individual’s biology reacts to those thoughts and experiences.\n\n## Words as Powerful Indicator of Mental Health: Designing an Experiment Around Language\n\nResearchers turned to language to probe participants’ internal emotional states in a controlled setting.\n\n“Language is deeply connected to the human condition,” said Byrd. Because mental health assessments often rely on clinician–patient conversations, the team used language as a window into emotional and cognitive processing.\n\nAt the core of the study was a sentence evaluation task in which participants read 160 self-referential statements, such as “I feel sad a lot of the time,” varying in emotional tone. While completing the task, participants’ brain activity, eye movements, and skin conductance responses were recorded simultaneously. After each statement, participants indicated whether they agreed or disagreed.\n\n## Mapping Neural Signatures of Depression and Suicidal Ideation\n\nResearchers used EEG to capture the brain’s millisecond-by-millisecond responses to emotionally charged language. Using a 64-channel EEG system, the team recorded electrical activity while participants read self-referential statements, allowing researchers to identify neural signatures associated with different mental health states.\n\nThe team found that these neural responses differed across healthy, depressed and suicidal groups. Using three computational approaches, researchers identified distinct patterns associated with depression and suicidal ideation. Jeong found that, based on time-resolved analyses of 64-channel EEG data, the most reliable differences between groups emerged 300 to 600 milliseconds after word presentation, a window associated with emotional and semantic processing.\n\nKommineni used deep learning models to analyze differences in how the brain responds to positive and negative sentences, finding that stronger differential responses helped distinguish healthy individuals from those with depression.\n\nMcDaniel used event-related potentials (ERPs), or direct brain responses to specific sensory or cognitive events, to isolate specific neural signals. The team found that the N400 response was reduced in depressed individuals when processing negative content and identified a blunting of the N170 response that appeared to track the severity of suicidal ideation and help distinguish suicide risk from general depression.\n\n## Tracking Eye Movements, an Involuntary Response, to Reduce Bias in Mental Health Assessments\n\nWhile brain signals provide a view into internal cognitive processing, the team also examined an equally involuntary physical response: how a person’s eyes move when processing emotional information.\n\nBecause gaze behavior is largely automatic and difficult to consciously control, eye movements can provide a more objective window into attentional and cognitive processing than traditional surveys.\n\nUsing an infrared SR Research EyeLink 1000 Plus eye-tracking system, researchers recorded participants’ eye movements as they completed the same sentence evaluation task used throughout the study. Researchers then applied a deep learning pipeline to identify patterns associated with depression and suicidal ideation.\n\nThe team found that horizontal gaze patterns carried the most diagnostic information, particularly when participants processed negative statements. While healthy participants tended to show more structured visual patterns aligned with the “disagree” option in response to negative content, individuals with suicidal ideation exhibited more dispersed or disengaged viewing patterns.\n\n## Hidden Signals in Sweat: Measuring Depression Through the Sympathetic Nervous System\n\nThe team also investigated whether electrodermal activity (EDA), a measure of physiological arousal through the body’s sympathetic nervous system, could serve as an objective biomarker for depression and suicidal ideation.\n\nUnlike self-reported symptoms, sweat responses occur automatically and are difficult to consciously control. Researchers recorded skin conductance responses as participants processed emotionally charged words and developed a computational model that analyzed the timing and intensity of those responses rather than averaging sweat activity across the entire experiment.\n\nThe team found that responses to negative words carried the strongest diagnostic information. Individuals with depression and suicidal ideation exhibited altered physiological reactions to negative content, suggesting that reactivity to emotionally distressing information may serve as a measurable biomarker for depression and suicidality.\n\n## A New Tool to Supplement Clinical Mental Health Assessments\n\n“Our goal isn’t to replace current mental health assessments but to supplement them,” said Byrd, emphasizing that the work is designed to add an additional layer to existing self-report surveys, helping psychiatrists make more informed, data-driven decisions and identify more precise intervention points for patients experiencing suicidal ideation.\n\nPublished on July 13th, 2026\n\nLast updated on July 13th, 2026", "url": "https://wpnews.pro/news/usc-study-used-sweat-brain-signals-eye-movements-and-ai-to-detect-depression-and", "canonical_source": "https://viterbischool.usc.edu/news/2026/07/usc-study-used-sweat-brain-signals-eye-movements-and-ai-to-detect-depression-and-suicide-risk/", "published_at": "2026-07-13 17:40:48+00:00", "updated_at": "2026-07-13 22:09:54.906525+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "ai-research", "ai-ethics", "ai-policy"], "entities": ["USC", "Shrikanth Narayanan", "DARPA", "PRECOG", "Kleanthis Avramidis", "Colin McDaniel", "Assal Habibi", "USC Signal Analysis and Interpretation Laboratory"], "alternates": {"html": "https://wpnews.pro/news/usc-study-used-sweat-brain-signals-eye-movements-and-ai-to-detect-depression-and", "markdown": "https://wpnews.pro/news/usc-study-used-sweat-brain-signals-eye-movements-and-ai-to-detect-depression-and.md", "text": "https://wpnews.pro/news/usc-study-used-sweat-brain-signals-eye-movements-and-ai-to-detect-depression-and.txt", "jsonld": "https://wpnews.pro/news/usc-study-used-sweat-brain-signals-eye-movements-and-ai-to-detect-depression-and.jsonld"}}