CMR-Analysis (including machine learning)

1346611 - Optimal method of measuring LV mass: an AI test-retest study

Wednesday, January 25, 2023

Location: E-POSTER

Presenting Author(s)

YJ

Yue Jiang, MSc

PhD student
University College London, United Kingdom

Primary Author(s)

YJ

Yue Jiang, MSc

PhD student
University College London, United Kingdom

Co-Author(s)

HS

Hunain Shiwani, MD

Clinical Research Fellow
University College London and Barts Heart Centre, United Kingdom
MA

Mohammed Alzahir, MD

Consultant Radiologist
Qatif Central Hospital, Saudi Arabia
AB

Anish N. Bhuva, MD, PhD

Consultant Cardiologist
Barts Health NHS Trust, England, United Kingdom
HX

Hui Xue, PhD

Director, Imaging AI Program
National Institutes of Health
Bethesda, Maryland, United States
Thomas A. Treibel, MD, PhD

Consultant Cardiologist
University College London, England, United Kingdom
Charlotte Manisty

Consultant Cardiologist
University College London and Barts Heart Centre
London, England, United Kingdom
PK

Peter Kellman, PhD

Senior Scientist
National Institutes of Health, Maryland, United States
AH

Alun Hughes, MD, PhD

Professor of Cardiovascular Physiology and Pharmacology
University College London, England, United Kingdom
James C. Moon, MD

Clinical Director, Imaging
Barts Heart Centre and UCL
London, England, United Kingdom
Rhodri Davies, MD, PhD

Associate clinical professor
University College London
London, Wales, United Kingdom

Disclosure(s):

Thomas A. Treibel, MD, PhD: No relevant disclosure to display

Charlotte Manisty: No relevant disclosure to display

Background:

Delineation of the left ventricular (LV) myocardium on Cardiac MR offers a reproducible method of measuring LV mass, but imprecise measurement introduces error.

LV mass can be measured at any stage of the cardiac cycle, and we aimed to see which point was most precise. We also compared the precision of LV mass measurement with detailed contouring of papillary muscles and trabeculae (trabeculated annotation) to rounded endocardial contours – see Figure 1.

Methods:

LV mass was measured at (1) diastole, (2) systole (3) the average of all phases (typically 25-30 phases) of the cardiac cycle. Measurement of LV mass using rounded endocardial contours was performed using a clinically validated machine learning model that amalgamates long axis views to determine the atrial-ventricular boundary [1]. The machine learning model was also retrained to produce detailed delineation of the papillary muscles and trabeculation [2].

Precision of LV mass measurement was evaluated using a multi-centre, multi-disease test-retest dataset of 102 patients [3]. The coefficient of variation (CoV) and limit of agreement (LoA) were used as indexes of measurement precision.

Results:

Coefficients of variation for all measurement methods are shown in the Figure 2 and Table 1. Results show that annotation in diastole leads to more precise LV measurement than measurement in systole (CoV 2.9% vs 5.9%, P< 0.001; LoA [-10.0, 10.4] vs [-21.8, 18.1]). There is no statistically significant difference in precision if LV mass measurement is made on all phases and averaged, or just in diastole (P=0.3). There is also no discernible difference in precision between rounded or trabeculated models.

Conclusion:

LV mass is best measured in diastole and should not be measured in systole. Precision is not improved when detailed annotation of the papillary muscles and trabeculation is performed.