AI-Powered Clinical Guideline Chatbots
(Nunez Lab)
These chatbots are AI-powered tools designed to guide users through evidence-based clinical guidelines for mental health and cancer care. They simplify complex medical information on conditions like bipolar disorder, depression, and cancer, offering clear, conversational support for diagnosis, treatment, and self-management. Each aims to improve accessibility, consistency, and understanding in patient and clinician decision-making.
Normative Digital Brain Twins Based on Connectivity and Brain Morphometry
Published in Lancet Digit Health in 2024 (Frangou)
This project develops normative “digital brain twins” derived from healthy populations using brain connectivity and morphometric features. Large-scale datasets of healthy individuals are used to construct age- and sex-adjusted reference models of structural and functional connectivity, cortical thickness, surface area, and subcortical volumes. For each patient, an individualized digital twin is generated by matching to the appropriate normative model, enabling precise quantification of deviations in connectivity and brain morphology. These deviation profiles are examined in relation to symptom dimensions, cognitive measures, and functional outcomes across psychiatric disorders. The approach emphasizes interpretability and robustness by relying on well-established neuroimaging measures rather than disease-trained classifiers. Key methodological goals include cross-site harmonization, reproducibility, and validation across independent cohorts. By integrating connectivity and morphometry within a normative digital twin framework, this project provides a principled and scalable strategy for characterizing neurobiological heterogeneity in psychiatry.
SanaMentis: AI-Powered Medical Documentation
(Bolaños, Tham, LeDue, and Murphy collaboration)
SanaMentis aims to streamline clinical documentation in psychiatry through innovative AI-driven tools, freeing clinicians to focus more fully on patient care.
📋 Digital Forms
This enables the efficiently gathering of patient data via secure, streamlined digital forms and patient portals.
🤖 AI Assistant
This AI assistant system will facilitate structured patient interviews and advanced data collection on these interviews through the use of generative AI. This project is currently in the development and prototype phase.
Synthetic Patient Interview Transcripts
Published in Frontiers in Digital Health in 2025 (Warner, Cao, LeDue, Tham, and Murphy collaboration)
Developing high-quality training data is essential for tailoring large language models to specialized applications like mental health. To address privacy and legal constraints associated with real patient data, we designed a synthetic patient and interview generation framework that can be tailored to regional patient demographics.
The patient profile creation process, based on official real-world statistics, ensures that each patient profile is diverse from the others and realistic, providing a strong foundation for generating dynamic interactions. Then, the transcript generation process can produce contextually rich interview transcripts, structured by a customizable question bank, with lexical diversity similar to normal human conversation. By integrating detailed patient generation with dynamic interviewing, the framework produces synthetic datasets that may aid the adoption and deployment of LLMs in mental health settings.
The project’s paper can be accessed here.
The CentileBrain Project
(Frangou)
CentileBrain is an AI-enabled neuroinformatics platform (https://centilebrain.org/) developed through global collaboration within the Lifespan Working Group of the ENIGMA Consortium The platform leverages large, harmonised multi-site datasets to learn normative trajectories of brain morphometry and connectivity across the lifespan. Using normative modelling, CentileBrain generates personalised centile scores and normative deviation scores that quantify deviation from expected patterns. It also supports brainAGE estimation, providing an interpretable summary of brain ageing relative to chronological age. CentileBrain is built to scale across cohorts and scanners, with transparent quality control, reproducible pipelines, and generalizability across diverse opulations.