# Personalized Prediction of Cardiovascular Outcomes through Machine Learning Analysis of Cardiac MRI and Genomics

> **NIH NIH R56** · YALE UNIVERSITY · 2024 · $753,813

## Abstract

Scientific Abstract
Clonal Hematopoiesis of Indeterminate Potential (CHIP) is a condition of aging that has been associated with
increased risk of heart disease and heart failure. CHIP is evaluated by DNA sequencing of peripheral blood.
However, not all patients with CHIP will develop heart failure or myocardial infarction. Our preliminary data
demonstrate that CHIP patients (cancer and non-cancer patients) have a higher burden of nonischemic scarring
in their hearts on cardiac MRI compared to patients without CHIP. In a pilot cohort using a machine learning
technique of convolutional neural networks (CNN), we are able to predict who have CHIP with 86% accuracy
and predicted future cardiomyopathy with an accuracy of 73%, thus opening the door to possible image guided
risk stratification for identifying patients with CHIP as well as predict future cardiac outcomes. We also find
through our novel deep learning algorithm that we can identify extended somatic variants that are CHIP-like that
can impact heart failure outcomes and through integration of patient data, genomic and imaging signatures, we
are able to develop partition risk scores that can weigh the variables such as CHIP and their contribution to heart
failure and the pathways involved. We hypothesize that using cardiac MRI, novel machine learning
techniques on large imaging datasets can identify CHIP vs non-CHIP patients as well as predict who will
develop adverse cardiovascular outcomes and that CNNs can be used on both imaging and genomic
signatures in large cohorts like the TOPMED, UKB and All of Us to predict cardiovascular outcomes
while pathway analyses will reveal mechanisms and novel targets for CHIP-mediated cardiomyopathy
Specific Aim #1 will use a novel CNN with multi-view cross-attention to identify MRI features that can accurately
predict CHIP as well as future adverse outcomes such as heart failure and myocardial infarction that will be
validated in large datasets like TOPMed, UKB and All of Us with a collective ~13000 cardiac MRIs. Specific Aim
#2 will use artificial intelligence approaches and a novel analysis pipeline that combines patient data, cardiac
MRI signatures and genomic signatures of CHIP and extended somatic variants that are predictive of
cardiovascular outcomes to help reveal mechanisms behind CHIP’s contribution to cardiomyopathy.

## Key facts

- **NIH application ID:** 11175780
- **Project number:** 1R56HL175627-01
- **Recipient organization:** YALE UNIVERSITY
- **Principal Investigator:** Jennifer M Kwan
- **Activity code:** R56 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $753,813
- **Award type:** 1
- **Project period:** 2024-09-17 → 2026-08-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/11175780

## Citation

> US National Institutes of Health, RePORTER application 11175780, Personalized Prediction of Cardiovascular Outcomes through Machine Learning Analysis of Cardiac MRI and Genomics (1R56HL175627-01). Retrieved via AI Analytics 2026-05-27 from https://api.ai-analytics.org/grant/nih/11175780. Licensed CC0.

---

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