# SCH: Efficient Image-based Hemodynamic Modeling via Physics-integrated Bayesian Deep Learning

> **NIH NIH R01** · UNIVERSITY OF NOTRE DAME · 2024 · $203,034

## Abstract

This project aims to revolutionize cardiovascular research and healthcare by developing an Al-augmented
image-based hemodynamic modeling platform. The primary objective is to establish efficient and reliable
data-enabled, patient-specific computational modeling capabilities to enhance the comprehensive
understanding of cardiovascular physiology and pathophysiology. It addresses the gaps in existing
image-based computational modeling frameworks that are labor-intensive, computationally expensive, and
subject to large uncertainties. Specifically, the project will automate the transformation of medical images
into precise 30 geometries for computational fluid dynamics (CFO) and fluid-structure interaction (FSI)
simulations via deep learning, significantly reducing manual labor and computational costs. By integrating
physics with mesh-based geometric deep learning through differentiable programming, this project aims to
enable fast surrogate CFD/FSI simulations, epitomizing an innovative blend of machine learning with
domain-specific knowledge and allowing for rapid predictions of functional information such as blood flow
patterns, pressure and wall shear stresses. Furthermore, an ensemble Bayesian learning framework will
be developed to propagate and quantify uncertainties in model predictions, thereby enhancing the
reliability and trustworthiness of the results. Finally, the introduction of advanced visual analytics to
interpret vast hemodynamic data ensembles underscores the project's commitment to accessibility and
user-centric design. Collectively, these innovations aim to enhance the accuracy, efficiency, and reliability
of patient-specific hemodynamic modeling, making them more accessible and actionable for healthcare
professionals. The project's long-term goals include improving the understanding, diagnosis, and treatment
of cardiovascular diseases, aligning with the National Heart, Lung, and Blood lnstitute's (NHLBI) mission to
combat heart, lung, and blood diseases and extend the lives of those afflicted. By offering a novel
approach to accurately and reliably model cardiovascular dynamics, the project holds the potential to
significantly impact the field of cardiovascular research and healthcare, providing a basis for better clinical
decisions and treatment strategies.
RELEVANCE (See instructions):
This research focuses on developing patient-specific models and diagnostic tools for cardiovascular
diseases, aligning with the NHLBl's goal of improving health and extending life by furthering the knowledge
and treatment of heart, lung, and blood conditions. By enhancing the accessibility and reliability of
image-based computational modeling, the project will revolutionize cardiovascular healthcare, significantly
advancing public health through improved prevention, diagnosis, and treatment.

## Key facts

- **NIH application ID:** 11062886
- **Project number:** 1R01HL177814-01
- **Recipient organization:** UNIVERSITY OF NOTRE DAME
- **Principal Investigator:** Jian-Xun Wang
- **Activity code:** R01 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $203,034
- **Award type:** 1
- **Project period:** 2024-09-01 → 2025-05-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/11062886

## Citation

> US National Institutes of Health, RePORTER application 11062886, SCH: Efficient Image-based Hemodynamic Modeling via Physics-integrated Bayesian Deep Learning (1R01HL177814-01). Retrieved via AI Analytics 2026-05-22 from https://api.ai-analytics.org/grant/nih/11062886. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*