Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/dataset.json

Name

FORCE registry cardiac MRI subset for single-ventricle ventricular segmentation

Link

https://www.forceregistry.org

Indexing

Keywords: Cardiac MRI, Single ventricle, Fontan, Short-axis cine, Ventricular segmentation, Deep learning, Ejection fraction, End-diastolic volume, End-systolic volume

Content: CA, MR, PD

RadLex: RID35976, RID12778, RID11250, RID10312, RID45645, RID1571, RID1564

SNOMED: 13213009

Author(s)

Tina Yao

Nicole St. Clair

Gabriel F. Miller

Adam L. Dorfman

Mark A. Fogel

Sunil Ghelani

Rajesh Krishnamurthy

Christopher Z. Lam

Michael Quail

Joshua D. Robinson

David Schidlow

Timothy C. Slesnick

Justin Weigand

Jennifer A. Steeden

Rahul H. Rathod

Vivek Muthurangu

Organization(s)

Boston Children’s Hospital

University College London

Fontan Outcomes Registry Using CMR Examinations (FORCE)

Children’s Hospital of Philadelphia

Nationwide Children’s Hospital

Hospital for Sick Children, Toronto

Ann & Robert H. Lurie Children’s Hospital of Chicago

Emory University School of Medicine

Texas Children’s Hospital

Funding

FORCE registry supported by a grant from the Additional Ventures Foundation. Study authors also report support from UKRI, NIH, British Heart Foundation (details in paper).

Ethical review

Multicenter retrospective study approved by the Boston Children’s Hospital IRB (IRB-P00028482) with waivers of consent; HIPAA compliant. Contributing institutions relied on Boston Children’s Hospital IRB or obtained local IRB/ethics approval with waivers of consent.

Comments

Multicenter, retrospective subset from the Fontan Outcomes Registry Using CMR Examinations (FORCE) used to develop and evaluate a deep learning pipeline for ventricular segmentation in patients with single ventricle physiology (Fontan circulation).

Date

Published: 2023-11-15

References

[1] Yao T, St. Clair N, Miller GF, et al.. "A Deep Learning Pipeline for Assessing Ventricular Volumes from a Cardiac MRI Registry of Patients with Single Ventricle Physiology". Radiology: Artificial Intelligence. 2024;6(1):e230132. 2023-11-15. doi:10.1148/ryai.230132. PMID: 38166332. PMCID: PMC10831511.

Dataset

Motivation

To enable standardized, automated core-lab style ventricular segmentation across the FORCE registry to reduce variability from heterogeneous clinical reports.

Sampling

Multicenter retrospective sampling across 13 institutions (training/validation/test) from the FORCE registry; external testing used 475 additional examinations including data from six sites not in training.

Partitioning scheme

250 complete CMR examinations randomly split 175/25/50 (training/validation/testing) ensuring patients with multiple scans were not split across sets; an additional 475 unseen examinations from 16 institutions used for external pipeline performance evaluation.

Missing information

Exact image resolutions, protocols, and per-series/image counts are not detailed in the manuscript.

Relationships between instances

Patients with multiple scans were kept within the same partition to avoid information leakage.

Noise

Heterogeneous image quality across centers; approximately 35% of short-axis stacks were subjectively suboptimal due to artifacts, poor contrast, or high noise.

Confidentiality

All examinations were de-identified on upload to the registry; HIPAA compliant.

Re-identification

Data were de-identified prior to analysis; no PHI present per manuscript.

Sensitive data

Includes pediatric and adult patients with congenital heart disease (single ventricle physiology).