How one company is creating a scalable biomarker for depression and anxiety
Selin Celikoyar • April 25, 2023
A third of the world’s population have symptoms of depression and anxiety. Mental health services were already over-stretched before 2020, but since the pandemic prevalence of anxiety and depression skyrocketed. Strict confinement measures, financial insecurity, and periods of intensifying COVID-19 deaths accelerated mental distress. Now imagine a world where this could be instantaneously measured like a blood pressure cuff, but instead of pulsatile beat of a blood vessel, the tone and sound wave of voice provides quantitative data. For Ellipsis Health, the time to revolutionize mental health screening is now! By recording a person’s voice answering questions for 2-3 minutes, deep machine learning via artificial intelligence (AI), and adding results to a growing national database, mental health screening could become more objective and scalable for consumers.
But why choose speech? Research has been done on speech and shows that it can point towards mental health and wellness. Patients with major depression often have speech that is slow, full of pauses, negative in content and lacking energy. Diagnostics using participants speech results in accuracy over 80%. Unlike filling out a survey, Ellipsis’ deep AI learning platform uses two machine learning algorithms: one that analyzes words using natural language processing (NLP) and other that analyzes acoustic properties of speech. NLP algorithm relies on automatic speech recognition (ASR) technology, which can account for issues like noise, accents, and stuttering while the NLP algorithm is trained to be reliable regardless of age, race, and dialect of the speaker. The clinically validated data has shown to be 20-45% more accurate at detecting depression symptoms than the current mainstream method of self-administering questionnaires. As a result of the current practice model, providers are in a difficult position between validating signs and symptoms of depression and appropriately treating patients in an acceptable timeframe.
By focusing on depression and anxiety first, Ellipsis believes that this technology will have the broadest and most meaningful impact first. For Ellipsis Co-Founder Mainul Mondal, working on these diagnosis first is personal because “I [have] seen personally with my family members… I thought that is a problem worth solving because it touches humanity.” For consumers, the modular technology can be plugged into mobile apps, patient portals, and incorporated into telehealth programs making it scalable solution, which has been validated by over 10,000 users to date. A recent non-randomized independent study focusing on seniors reaffirmed the NLP platforms feasibility with a 61% completion rate. Of the total cohort that participated, 30% of participants spoke longer than required by platform’s AI, suggesting a possible therapeutic pairing between the platform and user.
By utilizing AI to analyze speech semantics (words people say) and acoustics (how you say them) to create a clinically validated sign for depression and anxiety, Elipsis is hoping to solve mental health issues and increase independent wellness through measurement and equity. This not only revolutionizes the ability to first detect symptoms in patients but can also be used to check on current patients and see how they are maintaining. Especially in areas of with high rates of trauma and distress, NLP may help health practioners triage mental health care and keep track of survivors.
Lin, Nazreen, Rutowski, et al. Feasibility of Machine Learning-Based Smartphone Application in Detecting Depression and Anxiety in a Generally Senior Population. 2022. Frontier Psychology. Vol 13. https://doi.org/10.3389/fpsyg.2022.811517.
Learn more about: