Automated Deep-learning Right Ventricular 3D Strain Analysis from Routine 2D cine CMR: Feasibility and Prognostic Value in Dilated Cardiomyopathy

Thursday, February 5, 2026

6:20 PM - 6:30 PM

Location: Americas I and Hall 3 (2nd Lower Level)

Presenting Author(s)

BM

Benjamin A. Magyar

Biomedical Engineering Co-op Student
Libin Cardiovascular Institute, University of Calgary
Calgary, Alberta, Canada

Primary Author(s)

BM

Benjamin A. Magyar

Biomedical Engineering Co-op Student
Libin Cardiovascular Institute, University of Calgary
Calgary, Alberta, Canada

Co-Author(s)

RM

Rylan Marianchuk, BSc

Masters Student
University of Calgary
Calgary, Alberta, Canada
Justin J. Tse, PhD, MSc, BSc

Research Associate
University of Calgary
Calgary, Alberta, Canada
TB

Talia A. Beckie, BEng

Cardiac Imaging Systems Engineer
Libin cardiovascular Institute, University of Calgary
Calgary, Alberta, Canada
CJ

Cory R. Jensen, BSc

Cardiac Machine Learning Developer
Libin Cardiovascular Institute, University of Calgary, Canada
Ahmed Abdelhaleem, MD

Advanced Cardiac Imaging Fellow
West Virginia University Heart & Vascular Institute
Morgantown, West Virginia, United States
SD

Steven Dykstra, MSc, BEng

PhD Candidate
University of Calgary
Calgary, Alberta, Canada
FH

Fereshteh Hasanzadeh, MSc

Masters Student
University of Calgary
calgary, Alberta, Canada
JF

Jacqueline Flewitt, MSc

Manager of Strategic Partnerships
Libin Cardiovascular Institute; University of Calgary
Calgary, Alberta, Canada
YF

Yuanchao Feng, PhD

Senior AHS analyst
Libin Cardiovascular Institute; University of Calgary
Calgary, Alberta, Canada
JN

Julian B. Neves, MSc

Junior Data Engineer
Libin Cardiovascular Institute, University of Calgary, Canada
AA

Augustine Amakiri, PhD

Data Analyst
University of Calgary, Canada
MK

Melanie King, PhD

Director, Program Development and Implementation
Libin Cardiovascular Institute, University of Calgary, Canada
AH

Andrew G. Howarth, MD, PhD

Associate Professor
Libin Cardiovascular Institute; University of Calgary
Calgary, Alberta, Canada
CL

Carmen P. Lydell, MD

Radiologist
Stephenson Cardiac Imaging Centre, Libin Cardiovascular Institute of Alberta, University of Calgary
Calgary, Alberta, Canada
MB

Michael Bristow, MD

Clinical Associate Professor
University of Calgary, Alberta, Canada
MG

Marco J.W Götte, MD, PhD

MD, PhD
Stephenson Cardiac Imaging Centre
Calgary, Alberta, Canada
LK

Louis Kolman, MD

Clinical Assistant Professor
Libin Cardiovascular Institute; University of Calgary
Calgary, Alberta, Canada
NF

Nowell M M. Fine, MD

Associate Professor
Libin Cardiovascular Institute of Alberta, University of Calgary
Calgary, Alberta, Canada
RM

Robert JH Miller, MD

Cardiologist
Libin Cardiovascular Institute, University of Calgary, Canada
Dina Labib, MD, PhD, FSCMR

Associate Scientific Director, Personalized Diagnostics Program; Adjunct Assistant Professor
University of Calgary
Calgary, Alberta, Canada
James A. White, MD

Professor
Libin Cardiovascular Institute; University of Calgary
Calgary, Alberta, Canada

Background: Expanding evidence supports right ventricular (RV) health to provide meaningful contributions to future cardiovascular outcomes in patients with cardiomyopathy. This has led to interest in the use of CMR for phenotyping this complex structure to deliver reproducible markers of contractile health. 2D measures of RV deformation (i.e. strain) are typically limited by sparse representation using a single (4-chamber) plane. In this study we train and validate a fully automated, deep-learning-based pipeline for generation of 3-dimensional (3D) RV shape models across cardiac phases for the description of 3D RV strain. This technique was then validated in a previously unseen dataset of dilated cardiomyopathy (DCM) to assess its capacity to predict clinical outcomes.

Methods: A UNet-based segmentation algorithm was trained to isolate the RV borders on sequential short axis and long-axis cine images from 500 manually contoured studies across a wide range of disease phenotypes. Following training and internal validation, generated contours were provided to a custom pipeline to perform 3D image registration / breath-hold correction followed by standardized meshing of the RV. Meshes were comprised of spatially registered nodes organized into 456 tetrahedral elements, assigned for both free wall and septal regions. This pipeline was then used to conduct batch (unsupervised) analysis for a curated cohort of 514 patients with DCM completing CMR in the Cardiovascular Imaging Registry of Calgary (CIROC), all having LV dilation (BSA-indexed LVEDV >2SD above sex-specific references), LV EF < 50%, and absence of identifiable ischemic or non-ischemic cause. Free wall RV area strain (RVAS) was examined from the generated model and explored in a multi-variable model to determine associations with the composite outcome of heart failure and all-cause mortality.

Results: Baseline cohort characteristics are shown in Table 1. An example of an automatically generated 3D RV mesh model for a DCM patient is shown in Figure 1. Over a median follow-up of 5.1 years, 187 (36%) patients experienced the composite outcome. 3D RVAS was significantly associated with the composite outcome on univariable analysis. In a multivariable analysis, adjusted for age and sex, 3D RVAS remained predictive with a 1.5-fold increased risk per 1% worsening (HR 1.5 [95% CI 1.1-2.1], p=0.01). Patients with RVAS worse than a survival-based optimal cutpoint of -28% experienced a 2.51-fold increased risk of the composite outcome versus those below this threshold (HR 2.51, 95% CI 1.86-3.38, p< 0.001). Unadjusted Kaplan-Meier curves are shown in Figure 2.

Conclusion: Fully automated 3D RV strain analysis is feasible and was shown to deliver prognostically relevant markers in a large-scale, real-world MRI dataset of patients with DCM. Expanded investigation of this technique in a multi-centre setting, and across a broader range of cardiomyopathy states is required.

Baseline characteristics of the study cohort, stratified by the primary outcome

Characteristic	Overall N = 514	Outcome -ve, N = 327 (64%)	Outcome +ve, N = 187 (36%)	p-value²
Age, y	56.0 (45.0, 66.0)	57.0 (44.0, 65.0)	55.0 (45.0, 68.0)	0.82
Male sex	352 (68%)	220 (67%)	132 (71%)	0.44
Hypertension	228 (44%)	133 (41%)	95 (51%)	0.026
Diabetes mellitus	190 (37%)	115 (35%)	75 (40%)	0.26
Dyslipidemia	361 (70%)	221 (68%)	140 (75%)	0.082
Atrial fibrillation/flutter	90 (18%)	54 (17%)	36 (19%)	0.43
Chronic kidney disease	83 (17%)	47 (15%)	36 (19%)	0.23
CMR characteristics
LV EDV, mL/m²	129.8 (116.0, 149.4)	125.1 (113.0, 143.4)	141.1 (123.3, 161.8)	< 0.001
LV ESV, mL/m²	90.1 (70.8, 115.8)	78.4 (66.4, 104.1)	106.2 (85.3, 126.6)	< 0.001
LV EF, %	30.0 (21.8, 40.5)	35.1 (25.3, 43.2)	24.0 (18.8, 30.7)	< 0.001
RV EDV, mL/m²	94.1 (77.6, 114.8)	93.3 (77.5, 115.1)	96.2 (78.3, 111.7)	0.57
RV ESV, mL/m²	52.7 (37.4, 74.6)	49.7 (36.2, 70.5)	56.7 (42.7, 81.6)	0.002
RV EF, %	44.8 (30.9, 53.1)	47.2 (36.0, 54.4)	38.1 (27.0, 50.0)	< 0.001
Values are median (Q1, Q3), or counts (%)
CMR indicates cardiac magnetic resonance; LV, left ventricular; EF, ejection fraction; RV, right ventricular; BSA, body surface area; and LA, left atrial

Baseline characteristics of the study cohort, stratified by the primary outcome

Characteristic	Overall N = 514	Outcome -ve N = 327 (64%)	Outcome +ve N = 187 (36%)	p-value²
Age, y	56.0 (45.0, 66.0)	57.0 (44.0, 65.0)	55.0 (45.0, 68.0)	0.82
Male sex	352 (68%)	220 (67%)	132 (71%)	0.44
Hypertension	228 (44%)	133 (41%)	95 (51%)	0.026
Diabetes mellitus	190 (37%)	115 (35%)	75 (40%)	0.26
Dyslipidemia	361 (70%)	221 (68%)	140 (75%)	0.082
Atrial fibrillation/flutter	90 (18%)	54 (17%)	36 (19%)	0.43
Chronic kidney disease	83 (17%)	47 (15%)	36 (19%)	0.23
CMR characteristics
LV EDV, mL/m²	129.8 (116.0, 149.4)	125.1 (113.0, 143.4)	141.1 (123.3, 161.8)	< 0.001
LV ESV, mL/m²	90.1 (70.8, 115.8)	78.4 (66.4, 104.1)	106.2 (85.3, 126.6)	< 0.001
LV EF, %	30.0 (21.8, 40.5)	35.1 (25.3, 43.2)	24.0 (18.8, 30.7)	< 0.001
RV EDV, mL/m²	94.1 (77.6, 114.8)	93.3 (77.5, 115.1)	96.2 (78.3, 111.7)	0.57
RV ESV, mL/m²	52.7 (37.4, 74.6)	49.7 (36.2, 70.5)	56.7 (42.7, 81.6)	0.002
RV EF, %	44.8 (30.9, 53.1)	47.2 (36.0, 54.4)	38.1 (27.0, 50.0)	< 0.001
Values are median (Q1, Q3), or counts (%)
CMR indicates cardiac magnetic resonance; LV, left ventricular; EF, ejection fraction; RV, right ventricular; BSA, body surface area; and LA, left atrial

3D mesh reconstruction of the RV from routine 2D cine CMR at end diastole (top) and end systole (bottom). The model is shown in 3D (left) and overlaid on 4-chamber long-axis (LAX) and short-axis (SAX) cine images (right).

Kaplan Meier survival curve for freedom from the composite outcome for RV area strain dichotomized at the optimal cutpoint of -28%.