Skip to main content
Research topic

AI for Healthcare and Digital Health

Dr. Costilla Reyes’s work applying interpretable machine learning, behavioral sensing, digital phenotyping, and neurosymbolic methods to health.

From movement signals to health and security questions

Dr. Costilla Reyes’s doctoral work at the University of Manchester studied gait and footsteps as spatio-temporal signals. It connected floor-sensor measurements with two applications: gait changes associated with cognitive tasks, and footstep patterns used for biometric verification. The healthcare studies involved dozens of participants and the security studies more than a hundred.

Follow-on work compared temporal features and machine-learning models for gait-activity classification, and evaluated deep residual networks on floor-sensor footstep data. This research connects sensor design, representation learning, and pattern-recognition methods across healthcare and security applications.

Digital phenotyping and individual behavioral patterns

Digital phenotyping describes a person’s behavior in everyday life from data that personal devices already record. Later studies pair smartphone sensing with ecological momentary assessments, brief in-the-moment self-reports collected on a phone, to examine behavior in mental-health research. One schizophrenia study combined these data with individual-level network analysis; networks observed during changes such as visiting a different number of locations differed from baseline and randomly sampled periods.

A related study used population-level and personalized random-forest models to predict psychological-network measures, maps of how a person’s reported experiences relate to one another, from behavioral-sensing data. The publisher abstract reports stronger prediction of network measures than aggregated assessment scores. Together, these studies develop methods for measuring behavioral patterns over time and relating them to psychological-network measures.

Treatment-response modeling and explanations

In body dysmorphic disorder studies, the evaluated models examined response and remission in cognitive-behavioral therapy, including a smartphone-delivered trial. Treatment credibility, how believable participants find a treatment at the outset, repeatedly appeared among the useful predictors; decision trees exposed credibility thresholds that could be reviewed directly, and simpler models remained competitive with more complex ones.

A secondary analysis of a trial with several hundred participants across three smartphone-delivered interventions for major depressive disorder identified baseline anxiety, measured with the GAD-7 questionnaire, as the strongest predictor of treatment outcomes: higher anxiety scores went with lower odds of improvement. The study adds an interpretable view of factors associated with outcomes in digital mental-health research.

Health technology in practice

Dr. Costilla Reyes founded Equ Healthcare in April 2024 and leads it as founder and chief executive; since October 2025 he has also been director of research at Equ AI Lab. Equ Healthcare announced a collaboration with Mayo Clinic on October 28, 2025. He has discussed digital healthcare and explainable AI in health on panels including World Salon’s “From Lab to Cloud” (March 2025), the Congressional Hispanic Caucus Institute Tech Summit (November 2025), and a ZETA telehealth conversation (December 2025). Earlier, at McLean Hospital’s 2020 Technology in Psychiatry Summit, he presented his team’s open-source pipeline for multimodal digital-phenotyping data.

Writing places the research in a policy context: a 2021 chapter on the governance of AI in digital healthcare, which argues for multistakeholder partnerships, a long-form article on neurosymbolic AI in healthcare published in October 2025, and a November 2025 essay linking U.S. healthcare spending pressures to the case for transparent, explainable AI.

Research applications

Across gait sensing, digital phenotyping, and treatment-response modeling, Dr. Costilla Reyes’s research connects data collection with interpretable analyses that researchers and health experts can use to investigate movement, behavior, and treatment-response patterns.

Topics