Thoracic Digital Twin for Personalized Cardiorespiratory Modeling

Background

In cardiovascular healthcare, accurately monitoring deep internal cardiac mechanics and hemodynamics is vital for diagnosing and managing conditions ranging from progressive heart failure to congenital heart defects. While clinical imaging modalities such as four-dimensional computed tomography and cardiac magnetic resonance imaging provide high-resolution assessments of internal cardiac structures, they are expensive, time-consuming, and restricted to episodic hospital visits. Moreover, frequent imaging often exposes patients to cumulative radiation. Consequently, there is a critical need for continuous, non-invasive monitoring solutions—such as wearable mechanocardiographic sensors—capable of capturing real-time cardiorespiratory function outside clinical settings to enable proactive disease management.

However, translating surface-level wearable sensor data into actionable insights regarding internal cardiac mechanics remains a major challenge. As kinetic energy from the beating heart propagates toward the skin, it undergoes complex elastodynamic transmission, viscoelastic damping, and acoustic reflection across heterogeneous thoracic tissues, including the lungs, ribcage, musculature, and adipose layers. Current monitoring approaches fail to account for these patient-specific anatomical variations and structural propagation distortions. As a result, existing non-invasive wearables struggle to reliably correlate epidermal vibrations with deep intracardiac events, while purely computational biophysical models remain too computationally intensive for real-time clinical deployment.

Technology

The System for Thoracic Digital Twin Generation and Cardiorespiratory Prediction is a computational biophysics framework that constructs a patient-specific digital twin of the thoracic cavity. Initiating with baseline 4D medical imaging, the technology utilizes deep learning to segment thoracic anatomy and generate computational meshes. It then executes fluid-structure-electrophysiology interaction simulations to model intracardiac hemodynamics and myocardial contraction, calibrating these internal mechanics against imaging-derived heart wall motion. Next, a finite element model simulates how cardiac kinetic energy propagates through surrounding tissues to the skin surface across six degrees of freedom. In a second calibration loop, these predicted surface vibrations are optimized against real-time data from non-invasive wearable mechanocardiographic sensors, such as seismocardiography and gyrocardiography, creating a fully validated digital twin for continuous cardiac monitoring.

This solution is differentiated by its coupled dual-inverse calibration framework, which uniquely bridges deep internal cardiac biophysics with surface-level wearable sensor telemetry. Once initially calibrated, the digital twin updates longitudinally using wearable sensor streams alone, accurately tracking disease progression and myocardial remodeling without requiring repeated, radiation-based 4D imaging. To overcome the heavy computational burden of multi-physics solvers, the platform integrates physics-informed neural networks alongside a dual-tier cloud-edge architecture. The cloud generates a patient-specific predictive library of pathophysiological vibration signatures, enabling lightweight edge devices to detect anomalies in real time without continuous cloud connectivity. Additionally, the platform empowers clinicians to conduct in silico clinical trials, virtually simulating surgical or pharmacological interventions to predict post-procedural hemodynamic outcomes.

Benefits

• Enables non-invasive, continuous monitoring of deep internal cardiac mechanics using surface-bound wearable sensors
• Reduces the need for repeated, expensive, or radiation-based 4D medical imaging to track long-term disease progression
• Provides high diagnostic accuracy through personalized, patient-specific multi-physics modeling and dual-stage calibration
• Delivers near-real-time anomaly detection and risk assessment on local edge devices without requiring continuous cloud connectivity
• Facilitates virtual intervention planning and in silico clinical trials to predict patient responses to surgeries or pharmacotherapies
• Supports early, non-invasive detection of complex pathologies, including critical congenital heart defects in pediatric and neonatal patients

Applications

• Remote heart failure progression monitoring
• Pre-surgical cardiovascular intervention planning
• Neonatal congenital heart defect screening
• In silico clinical drug trials
• Wearable sensor diagnostic software integration

Additional info

Dr. Amirtahà Taebi is an Assistant Professor with the Department of Bioengineering at Lehigh University. His work focuses on developing AI-powered sensing technologies, wearable and contactless monitoring systems, and digital twin modeling to advance noninvasive cardiovascular diagnostics. His research has been supported by local, state, and federal agencies, including the NSF, NIH, and Mississippi Institutions of Higher Learning.