Highlighted Projects

AI-Facilitated ECG Analysis for Cardiovascular Prediction

This initiative advances cardiovascular care by using AI to extract deeper insights from standard electrocardiograms (ECGs). Our sophisticated AI models analyze subtle patterns in ECG data to predict a patient’s future risk for critical cardiac events, such as heart attacks, heart failure, and stroke. Our innovative approach moves beyond traditional risk factors, enabling highly personalized preventive strategies.

Wearable Perfusion Monitoring Device

This project introduces a new, AI-integrated wearable device for continuous, objective assessment of blood flow to tissues following surgery. Utilizing advanced sensor technology, the device employs AI algorithms to analyze real-time blood flow data, promptly detecting early signs of compromised circulation. This immediate alert system enables quick intervention, significantly reducing risks such as artery disease and amputation.

AI-Powered Computer Vision Tools for Surgical Assessment

This project modernizes surgical education by introducing AI-powered computer vision tools for objective performance evaluation. Employing advanced machine learning, the system analyzes video recordings of surgical procedures to precisely track instrument motion and assess skills such as efficiency and precision. This automated approach provides trainees with detailed, objective feedback, accelerating skill development and enhancing surgical proficiency.

AI-Facilitated Opportunistic Risk Stratification Using Routine CTs

This project leverages AI to unlock hidden health insights from routine Computed Tomography (CT) scans performed for other indications. Our AI algorithms opportunistically analyze these existing images to detect subtle, often overlooked, indicators of various underlying health risks, such as osteoporosis, cardiovascular disease or sarcopenia. This analysis enables proactive identification of at-risk individuals, facilitating early intervention without additional radiation exposure.

Treatment Estimation in Patients with Presumed Bacterial Infections

This project develops sophisticated counterfactual networks applying large-scale observational data to estimate causal pathways for treatment effectiveness in patients with presumed bacterial infections. The AI models aim to predict the optimal treatment outcome for an individual patient under different treatment scenarios. This provides doctors with data-driven insights to refine antibiotic prescribing, minimizing resistance and improving patient-specific outcomes.

Identifying Novel Parameters in Echocardiography to Assess Right Ventricular Function Using Machine Learning

This project uses machine learning techniques to uncover novel echocardiographic parameters that more accurately assess right ventricular (RV) function. By analyzing a large dataset of heart ultrasound images, the team aims to improve diagnostic accuracy for RV dysfunction—an area where traditional metrics fall short. The goal is to support cardiologists with new, data-driven tools to evaluate heart health and guide treatment decisions earlier and more effectively.

Machine Learning Analysis of Right Ventricular Shape for Characterization of Patients with Functional Tricuspid Regurgitation

This project applies machine learning to 3D imaging data to analyze subtle variations in right ventricular (RV) shape among patients with functional tricuspid regurgitation (FTR). By identifying structural biomarkers associated with FTR severity, the team aims to improve classification of disease subtypes and inform individualized treatment strategies. The approach advances the use of computational imaging for precision cardiology.

Leveraging Radiomic and Genomic Markers for Precision Diagnosis of Cardiac Sarcoidosis

This project integrates radiomic features from cardiac imaging with genomic data to improve the diagnosis of cardiac sarcoidosis—a rare and challenging inflammatory heart condition. By combining high-dimensional data sources, the team aims to develop machine learning models that enable earlier and more accurate detection. This precision approach could support more targeted therapies and better patient outcomes.

Digital Twinning of Medical Records and Exposome

This project creates “digital twins” of patients by combining electronic medical records with exposome data—factors like environment, lifestyle, and social determinants of health. Using advanced AI and modeling techniques, the goal is to simulate patient trajectories and predict long-term health outcomes. These digital replicas support proactive, personalized care and help uncover how external factors shape cardiovascular health over time.

AI–Computer Vision on Echocardiography for Evaluating Cardiac Structure and Function for Precision Phenotyping

This project applies advanced computer vision techniques to echocardiograms to extract detailed, quantifiable features of cardiac structure and function. The AI-driven approach aims to enable precision phenotyping—identifying nuanced subtypes of heart disease that are often missed with traditional analysis. The goal is to support more tailored diagnosis and treatment strategies in cardiovascular care.

Generative AI to Reduce Radiation and Enable Personalized Metabolic Phenotyping

This project leverages generative AI models to reconstruct high-quality metabolic imaging from low-radiation scans, reducing patient exposure while preserving diagnostic accuracy. By enhancing image clarity and extracting precise metabolic features, the approach enables more personalized phenotyping of cardiovascular conditions. The goal is to make advanced imaging safer and more accessible for long-term disease monitoring and prevention.

Transcranial Magnetic Stimulation

This project explores the use of transcranial magnetic stimulation (TMS) to modulate brain activity for therapeutic applications. Combining advanced imaging and stimulation techniques, the team aims to better understand and optimize TMS protocols for psychiatric conditions. The research seeks to enhance treatment efficacy and patient outcomes through personalized neuromodulation strategies.

Personalized Insulin to Carbohydrate Ratio for Individuals with Type 1 Diabetes

This project focuses on developing personalized insulin dosing algorithms to optimize the insulin-to-carbohydrate ratio for individuals with type 1 diabetes. Using patient data and AI models, the team aims to improve blood sugar control and reduce the risk of hypo- and hyperglycemia. The goal is to empower patients with tailored insulin management strategies for better daily glucose regulation.

Photoacoustic Tomography for Early Detection of MASLD

This project brings a non-invasive imaging approach to liver care by using photoacoustic tomography (PAT)—a hybrid light-and-sound modality—to visualize deep-tissue markers of metabolic dysfunction–associated steatotic liver disease (MASLD). By quantifying vascular architecture, lipid content, and speed-of-sound–derived tissue properties in validated preclinical models, the team aims to detect disease earlier and lay the groundwork to lessen dependence on biopsy while improving cardio-metabolic risk management.

Predicting Adherence To Home Rehabilitation With A Smartwatch

 This project applies machine learning to smartwatch data to predict adherence to a home rehabilitation program for frail end-stage kidney disease patients. Through obtaining accurate predictions of how long and what type of exercises patients are doing, exercise prescriptions can be better tailored towards individuals’ needs and goals. This will improve the likelihood of patients obtaining kidney transplants.