AI-based system detects depression from involuntary body reactions

A multidisciplinary team from the University of Southern California, including engineers, neuroscientists, linguists, and psychiatrists, created the PRECOG project to address the limitations of current psychiatric diagnostics, which rely on subjective questionnaires and self-reports. In experiments, volunteers read 160 emotionally varied statements and indicated agreement or disagreement while their physiological responses were continuously recorded via three types of sensors: a 64-channel EEG measuring brain electrical activity, an infrared eye tracker monitoring gaze, and skin conductance measurements to assess sweat composition. Deep neural networks analyzed the collected data and identified stable patterns associated with depressive disorders. EEG decoding showed that the most pronounced anomalies in processing negative words occur in the brain between 300 and 600 milliseconds after reading them, while gaze analysis revealed that healthy individuals display clear horizontal eye movement patterns toward the 'disagree' option when reading negative phrases, whereas depressed patients show chaotic, scattered, or inhibited movements. Mathematical processing of skin conductance data also detected altered autonomic responses to stressful words, related to sympathetic nervous system activity. The new AI platform is intended to serve as a support tool for practicing psychiatrists, potentially improving prediction of crisis states based on precise biological parameters. The information is purely educational, based on scientific research, and is not a personal medical recommendation.