Time Series

Comparing aggregate and individual-level (N-of-1) analyses of Oura Ring and smartphone survey data from 256 pregnant participants in the BUMP study shows that population-level trends often fail to reflect individual pregnancy trajectories, supporting personalized wearable-based monitoring.

A graph neural network that forecasts third-trimester pregnancy symptoms (edema and fatigue) from wearable physiological data and daily app surveys, with heart rate variability and low-intensity activity emerging as key predictors.

SPARC, a transformer-based foundation model pretrained on more than 100 million longitudinal laboratory and vital-sign measurements from 224,047 patients with cancer, used to predict treatment toxicities, adverse events and mortality.

We designed a stimulus-based heart rate variability protocol (active standing, deep breathing and stomach distention) and used logistic regression to weigh what each test adds to assessing autonomic function in chronic constipation.

Retrieval-grounded LLM agents that let researchers query large wearable health datasets in natural language, with privacy preserved by restricting the model to controlled SQL and Python execution layers instead of raw data access.

Interpretable change point detection method

This study finds that machine learning models trained on pooled electronic health record (EHR) data from multiple sites can generalize better to unseen data, while also revealing that data sharing partnerships may not always enhance performance, and certain baseline models continue to be competitive except in few-shot learning scenarios.

This research project aims to use data from wearable devices to assess delivery readiness during pregnancy on an individual basis. The study collects physiological data from pregnant women to identify specific maternal indicators that could predict delivery readiness and potentially reduce the risk of adverse outcomes for both mothers and infants.