# Reducing cardiac toxicity with deep learning and MRI-guided radiation therapy

> **NIH NIH R01** · UNIVERSITY OF WISCONSIN-MADISON · 2021 · $524,106

## Abstract

Cardiac toxicity is a devastating complication of cancer treatment and occurs during, shortly after, or even many
years after treatment. Long-term follow up of patients undergoing thoracic radiation, such as lymphoma, lung,
and esophageal cancers, has shown that in particular, radiation therapy (RT) can lead to major radiation-induced
cardiac toxicities like congestive heart failure and coronary artery disease. Typically, the standard of care for
cardiac dose assessment involves simple heart dose/volume metrics. However, mounting evidence suggests
that cardiac substructures contained within the heart are highly radiosensitive and dose to substructures
are more strongly associated with overall survival than assessing whole-heart dose/volume metrics.
Nevertheless, precise characterization of cardiac substructure dose in routine clinical practice is currently
limited because substructures are not visible on CT simulation scans used for RT planning, cardiac MRI are
not widely available for cancer patients, and manual delineation is cumbersome, taking 6-10 hours per case.
Further, precise localization is complicated by both cardiac and respiratory motion. Our long-term goal is to
develop and validate clinically viable novel technologies to localize cardiac substructures for novel cancer
therapies and interventions. The rationale for the proposed research is that by developing a robust and efficient
clinical framework for cardiac substructure dose assessment, more effective cardiac sparing strategies can be
achieved. Our expertise in deep learning coupled with experience in MR-guided RT has laid the groundwork for
this paradigm-changing proposal with the long-term goal of optimal cardiac sparing to ultimately reduce radiation-
induced cardiac toxicity. To attain the overall objectives, we propose the following specific aims: (i) develop
high quality, efficient cardiac substructure segmentation and accurate synthetic CT generation via deep learning,
(ii) quantify respiratory and cardiac-induced cardiac substructure motion using a novel 5D-MRI approach and
inter-fraction uncertainties to derive margins and planning strategies for robust cardiac sparing, and (iii) evaluate
the clinical efficacy of these emerging technologies in a randomized clinical trial for lung cancer evaluating
longitudinal changes in cardiac function from MRI, quality of life, echocardiogram, and blood biomarkers between
MR-guided adaptive radiation therapy with sparing and standard x-ray based treatment with whole-heart dose
metrics. This multi-disciplinary (oncology, cardiology, radiology, and computer science) proposal integrates state
of the art technologies while challenging the standard of care of using whole-heart dose evaluations. The
research proposed is innovative as it challenges the current, oversimplified classic model of whole-heart dose
estimates via several cutting-edge techniques. The research is significant because of its widespread application
in other thoracic can...

## Key facts

- **NIH application ID:** 10299368
- **Project number:** 1R01HL153720-01A1
- **Recipient organization:** UNIVERSITY OF WISCONSIN-MADISON
- **Principal Investigator:** Carri Kaye Glide-Hurst
- **Activity code:** R01 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2021
- **Award amount:** $524,106
- **Award type:** 1
- **Project period:** 2021-08-23 → 2026-07-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10299368

## Citation

> US National Institutes of Health, RePORTER application 10299368, Reducing cardiac toxicity with deep learning and MRI-guided radiation therapy (1R01HL153720-01A1). Retrieved via AI Analytics 2026-05-23 from https://api.ai-analytics.org/grant/nih/10299368. Licensed CC0.

---

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