AI Behavioral Health Project


Depression and anxiety are the two most common mental health disorders worldwide, yet a large proportion of those suffering do not receive treatment due to a lack of widespread screening and accurate diagnostic tools. Current methods for diagnosing depression and anxiety include self-reported patient surveys along with in-person clinical interviews to assess symptom presence and severity. These are highly subjective and resource-intensive methods. Consequently, there is a critical need for objective, automated tools that facilitate widespread screening of depression and anxiety.

Due to the subtle behavioral changes that manifest in patients with depression and anxiety, in visual, auditory, and linguistic domains, the problem of depression and anxiety diagnosis lends itself well to a digital, artificial intelligence--based approach. Spanning natural language processing, computer vision, and machine learning for audio analysis, AI-based approaches may be leveraged to detect these nuanced behavioral changes. As a result, the central question of this research is: Can AI methods detect the presence and severity of depression and anxiety?

To explore this question, we collected a novel dataset at the Stanford Family Medicine Clinic. This dataset consists of audio recordings, video recordings, and text transcriptions of casual conversations, like those occurring during intake and rooming at the beginning of a clinic visit. We estimated the presence and severity of depression and anxiety using unimodal and multimodal deep learning models that take as input these audio, video, and text data. We then used saliency analysis to determine the most significant visual and textual features, to illuminate behavioral and digital markers of depression and anxiety that our models use.

We proposed methods of integrating these novel screening tools into clinical care pathways to expand screening for common mental health disorders. This research introduced new AI models for automated mental health screening, generated meaningful clinical insights on behavioral markers of depression and anxiety within our dataset, and charted a path toward AI-enabled, widespread, accessible mental health screening.



Dr. Nirav Shah

CERC Senior Scholar

Dr. Nirav Shah is a Senior Scholar at Stanford University’s Clinical Excellence Research Center. He is a leader in patient safety and quality, innovation and digital health, and the strategies required to


transistion to lower-cost, patient-centered health care. Board-certified in Internal Medicine, Dr. Shah is a graduate of Harvard College and Yale School of Medicine, and is an elected member of the National Academy of Medicine. 

Dr. Shah serves as an independent director for STERIS plc, as an Advisor to Deerfield Management, and as a trustee of The John A. Hartford Foundation. He is a Senior Fellow of the Institute for Healthcare Improvement (IHI), and helps set the health priorities for the United States as a member of the HHS Secretary's Advisory Committee on National Health Promotion and Disease Prevention Objectives for 2030. Previously, he served as senior vice president and Chief Operating Officer for clinical operations for Kaiser Permanente in Southern California, and as Commissioner of the New York State Department of Health.


Tracy Terada

Research Operations Manager

Tracy is a 15+ year adminstrative veteran for the Stanford School of Medicine.  She started at the Lane Medical Library and is currently with the Clinical Excellence Research Center. 


Samira Daswani

Affiliated Scholar

Samira Daswani, is an entrepreneur working at the intersection of design, healthcare and technology. Using the power of human-centered design, she has tackled some of the most pressing problems of our generation - oncology, mental health, isolation and burnout.  


Ms. Daswani is currently the VP of Product at Visby Medical - a diagnostics company that designed and launched the first single-use PCR device. The first 2 products are for covid-19 and sexual health. Over the past decade, Samira has built a portfolio of products and companies that she has launched. As a part of Accretive, Samira played an instrumental role in the launch of 2 separate companies across the fashion and healthcare industry. While at Stanford, together with a group of talented individuals, Samira co-founded a venture that helps individuals better manage their emotional wellbeing. Her career started as a strategy consultant at McKinsey & Company. She earned a Master's degree in Design at Stanford University, a bachelor’s degree in biological engineering and art history from MIT and Wellesley College respectively.

On the personal side, Samira is a breast cancer survivor. Outside of giving back to the cancer community, she is an avid scuba diver, a red belt in taekwondo, and enjoys sketching on the weekends.


Stanford Healthcare AI Applied Research Team (HEA3RT)

Dr. Amelia Louise Sattler

Clinical Assistant Professor, Associate Director

Dr. Sattler is the Associate Director of HEA3RT. She joined Stanford Family Medicine in 2013, and is an ebullient family physician with special interests


in quality improvement, population health, medical education, adolescent medicine and mental health. In addition to her role with HEA3RT, Dr. Sattler is the Program Director for the Stanford Primary Care Performance Enhancement Program (PC-PEP), a rapid-cycle quality improvement program that empowers front-line faculty and staff to tackle day-to-day operational problems. She is also the Quality Improvement Lead for both the Stanford Primary Care Faculty Practices and the Stanford School of Medicine Continuity of Care Clerkship.


Margaret Smith MBA

Director of Operations

Margaret Smith is the Director of Operations of HEA3RT where she works with industry collaborators, and clinical and operational leaders


at Stanford on the development and implementation of artificial intelligence technologies that improve the lives of patients, providers and health systems. Her expertise lies in healthcare quality improvement, complex problem solving, facilitating cross discipline collaboration, and design thinking. Margaret holds a bachelor’s degree in finance and risk management, a master’s in business administration with a specialization in healthcare management from Baylor University, Robbins Institute for Health Policy and Leadership, and a Lean Six-Sigma black belt certification.



Grace Eunhae Hong

Research Associate

Grace Hong was born and raised in Illinois and graduated from Stanford University in 2019 with a B.A. in Economics.


Grace works with the Stanford Healthcare AI Applied Research Team (HEA3RT) to study the implementation of AI technologies in healthcare settings and has a special interest in better understanding how innovations in health technology can be used to improve access to healthcare and remedy health disparities.



Jeannie Yejin Jeong

Research Associate

Jeannie Jeong was born and raised in Yong-in, South Korea, and graduated from Vanderbilt University in 2022 with a B.A. in Psychology and Cognitive Science.


As a member of HEA3RT, Jeannie hopes to study and support the implementation of human-centered AI in healthcare to construct an improved and more equitable health system. Jeannie is particularly interested in integrating AI into mental healthcare.





Department of Computer Science

Dr. Ehsan Adeli

Scientist, Stanford AI Lab, Stanford Vision and Learning, Computer Science Department
Clinical Assistant Professor, Department of Psychiatry and Behavioral Sciences, Stanford School of Medicine

Neha Srivastha

Graduate Student - Joined  2021    

Neha Srivathsa is a 2023 first-year PhD student in Computer Science and is currently rotating with Prof. Fei-Fei Li. Her research interests are in machine learning for clinical medicine.