Research

Research Areas

  • Develop cutting-edge tools for the accessible and automated detection of cardiovascular diseases using widely available diagnostic modalities. Learn more.

  • Develop tools that tailor the assessment of randomized clinical trials through advanced data science and machine learning techniques. Learn more.

  • Utilize advanced computational methods on electronic healthcare records to assess and enhance the quality of healthcare delivery. Learn more.

  • Establish methodological best practices, assess adherence to evidence-based practices and investigate health-related outcomes to advance cardiovascular health. Learn more.

  • Develop digital twin technology to enhance personalized medicine and translate clinical trial inference to real-world care. Learn more.

Advanced cardiovascular diagnostics

We are developing state-of-the-art tools for the accessible and automated detection of cardiovascular diseases from ubiquitously available diagnostic modalities. Our deep learning models have demonstrated the ability to accurately detect cardiovascular conditions using electrocardiogram (ECG) images, portable single-lead ECGs, and heart ultra-sonograms. To bridge the gap from bytes to bedside, following robust model development and validation, we are assessing their performance in real-world settings and the feasibility of their seamless integration into clinical workflows.

Precision clinical trial inference

We are conducting a series of investigations on the development of tools that personalize the assessment of randomized clinical trials through the application of advanced data science and machine learning. These evidence-based tools extract individualized treatment effects based on participant-level data and promote the use of shared decision-making through personalized inference.

  • Individualising intensive systolic BP reduction in HTN using computational trial phenomaps and machine learning

    The Lancet Digital Health, Nov 22

  • Phenomapping-Derived Tool to Individualize the Effect of Canagliflozin on Cardiovascular Risk in Type 2 Diabetes

    Diabetes Care, Apr 2022

  • A phenomapping-derived tool to personalize the selection of anatomical vs. functional testing in evaluating chest pain

    EHJ: July 7, 2021

Electronic healthcare quality inference

Our lab conducts interdisciplinary research with a focus on utilizing advanced computational methods to enhance the quality of healthcare delivery. By employing natural language processing techniques, we extract meaningful clinical information from unstructured text in electronic health records to support decision-making and quality assessment. We deploy federated learning strategies to collaborate across multiple institutions, while respecting patient privacy and data security.

  • Multinational patterns of second line antihyperglycaemic drug initiation across CV risk groups: federated pharmacoepidemiological evaluation in LEGEND-T2DM

    BMJ: Oct 6, 2023

  • Automated Identification of Heart Failure with Reduced Ejection Fraction (HFrEF) using Deep Learning-based Natural Language Processing Techniques

    JACC: Heart Failure, Oct 23, 2024

  • Retrieval-Augmented Generation for Extracting CHA₂DS₂-VASc Risk Factors from Unstructured Clinical Notes in Patients with Atrial Fibrillation

    Medrxiv, Sep 22, 2024

Health outcomes and policy

In large national registries and datasets, we have defined both methodologic best practices for rigorous investigation and identified adherence to evidence-based cardiovascular care. Further, our lab is conducting research using national databases to investigate health-related outcomes and to inform health policy decisions that promote equitable access to healthcare, optimize resource allocation, and ultimately enhance the health and well-being of individuals and communities.

Digital twin technology

We are building digital twins of randomized clinical trials (RCTs) to translate their inference to real world clinical care. Our technology uses generative AI to simulate real world patient populations and translate clinical inference to these populations from RCTs and has demonstrated this translation across RCTs. We also incorporate multi-modal data processed by deep learning models to build a digital-twin signature of patient populations and improve causal inference. Our digital twin model of RCTs are currently under validation across pairs of RCTs.

  • Cardiovascular care with digital twin technology in the era of generative artificial intelligence

    EHJ, Sep 26, 2024

  • A Novel Digital Twin Strategy to Examine the Implications of Randomized Clinical Trials for Real-World Populations

    MedRxiv, Sep 26, 2024