CMR-Analysis (including machine learning)

1351683 - Smooth and trabeculated left ventricular reference ranges on 4405 healthy subjects; integrating AI analysis with in-line deployment

Wednesday, January 25, 2023

Location: E-POSTER

Presenting Author(s)

HS

Hunain Shiwani, MD

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

Primary Author(s)

HS

Hunain Shiwani, MD

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

Co-Author(s)

Rhodri Davies, MD, PhD

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

Jessica Artico, MD

Clinical Research Fellow
St Bartholomew's Hospital, England, United Kingdom
MW

Matthew Webber, MD, BSc

Clinical Research Fellow
University College London, United Kingdom
GJ

George Joy, MBBS

Research Fellow
University College London, United Kingdom
Joao B. Augusto, MD

Consultant Cardiologist
Hospital Prof Doutor Fernando Fonseca
Lisbon, Portugal
RH

Rebecca Hughes, MD

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

Stefania Rosmini, MD

Consultant Cardiologist
King's College Hospital NHS Foundation Trust, England, United Kingdom
IP

Iain Pierce, PhD

Scientist
Barts Heart Centre at St Bartholomew's Hospital, United Kingdom
Joao Cavalcante, MD

Director, Cardiac MRI and Structural CT Labs
Minneapolis Heart Institute at Abbott Northwestern Hospital
Minneapolis, Minnesota, 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
HX

Hui Xue, PhD

Director, Imaging AI Program
National Institutes of Health
Bethesda, Maryland, United States
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
Gabriella Captur, n/a

Consultant Cardiologist, Senior Clinical Lecturer
UCL
London, England, United Kingdom

Disclosure(s):

Joao Cavalcante, MD: No relevant disclosure to display

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

Charlotte Manisty: No relevant disclosure to display

Background: Cardiovascular magnetic resonance (CMR) quantitative analysis of left ventricular (LV) structure and function discriminates health from disease - provided the normal range is known. Published reference intervals cover a limited age range, do not reflect real-world body weight, depend on expert manual contours and are not method-specific (e.g., smooth vs trabeculated contours). We applied automated analysis to a large cohort of healthy volunteers and deployed the resultant reference ranges onto the scanner in-line to avoid any post-processing.

Methods:

CMR cines from 4118 healthy reference UK Biobank participants (age mean+/- SD 61.4 +/- 7.3) (range 46-80) and 491 locally recruited healthy volunteers (age mean+/- SD 40.3 +/- 12.9) (range 20-88) were combined to generate a reference cohort spanning 7 decades. Participants of real-world body weight were included (body mass index >14 and < 36kg/m²). All scans were automatically analysed in-line using a clinically proven AI algorithm (1,2) to calculate end-diastolic volume (EDV), end-systolic volume (ESV), LV mass (LVM) and ejection fraction (EF) with smooth and trabeculated contours [Figure 1]. Outliers >3 interquartile ranges below the 1^st quartile or above the 3^rd quartile were excluded leaving 4405 reference subjects. Each metric was indexed to body surface area (BSA, kg/m²), height (m) or height squared (m²). Linear regression was performed for each metric (non-indexed and indexed) against age and separately for each sex. Reference intervals were then deployed in-line into the scanner.

Results: Separate sex-specific regression formulas were calculated for the mean, +95% confidence interval, and -95% confidence interval. These are displayed in Figure 2 with a real-life patient example superimposed. In-line deployment of measurements and patient-specific reference ranges for smooth and trabeculated contour segmentations, (non-indexed and indexed) are provided in Figure 3.

Conclusion:

We report reference intervals for AI analysis of the LV using both smooth and trabeculated segmentation models deployed in-line on the scanners. At the time of submission, these authoritative, real-world, and large-scale reference intervals are already in clinical use across 10 scanners in major tertiary centres in the UK and USA scanning over 1000 patients per month.