A United States-based Nigerian scholar, Prof. Nchebe-Jah Iloanusi, has developed an Artificial Intelligence (AI) model capable of predicting the onset of depression in real time by analysing signals from the heart, sleep patterns, and speech biomarkers.
The breakthrough, unveiled by Iloanusi of the Department of Biology, College of Staten Island, represents a significant advance in the early detection and management of mental health conditions.
The multimodal system combines Heart Rate Variability (HRV), longitudinal sleep data, and speech-based features extracted with the aid of Large Language Models (LLMs). By fusing these signals, the AI can detect subtle shifts in physiological and behavioural states that often appear days or weeks before depression becomes clinically apparent.
According to Iloanusi, depression affects about 280 million people globally, yet an estimated 76 per cent remain undiagnosed until severe symptoms emerge. “We present the first clinically validated system that integrates wearable physiological sensing and speech-based behavioural biomarkers, processed through a transformer-based AI architecture, to predict depression onset in real time,” he said.
He noted that the system addresses key weaknesses of traditional screening, which depends on self-reporting and episodic clinical evaluations prone to recall bias, underreporting, and diagnostic delays. Instead, continuous physiological metrics such as HRV and sleep architecture, combined with speech analysis of tone, rhythm, and syntax, offer a more reliable foundation for real-time mental state inference.
Experts believe that if validated in larger populations, the model could anchor next-generation digital health platforms, capable not only of detecting depression but also predicting relapses and treatment responses. Embedding the technology into wearables, smartphones, and telehealth systems could allow healthcare providers to monitor patients remotely and intervene early when warning signs emerge.
“Depression is one of the most widespread mental health disorders globally, and delayed diagnosis often leads to worse outcomes, including chronic illness and suicide,” Iloanusi explained.
“This system bridges a critical gap by enabling proactive detection before clinical thresholds are reached.”
The model’s fusion-based design is particularly notable: HRV metrics reflect stress and autonomic nervous system activity; sleep quality and duration are closely tied to mood stability; and speech biomarkers reveal cognitive and emotional shifts. Together, they improve predictive accuracy and reduce false positives, outperforming reliance on single data streams.
While the research shows promise, Iloanusi stressed that the tool is not a substitute for psychiatric evaluation. He highlighted the need for ethical safeguards, strict data privacy measures, and large-scale clinical trials before deployment.
“AI can empower clinicians and patients, but it must be deployed responsibly,” he said, underscoring the importance of transparency and patient consent in digital mental health innovations.
