Rapid, Efficient Imaging

1351864 - Fully Automated Contrast Selection of Bright- and Black-blood LGE Images for Robust Myocardial Scar Imaging

Wednesday, January 25, 2023

Location: E-POSTER

Presenting Author(s)

Vd

Victor de Villedon de Naide, MSc

MSc
IHU LIRYC, Electrophysiology and Heart Modeling Institute, Université de Bordeaux – INSERM U1045, France

Primary Author(s)

Vd

Victor de Villedon de Naide, MSc

MSc
IHU LIRYC, Electrophysiology and Heart Modeling Institute, Université de Bordeaux – INSERM U1045, France

Co-Author(s)

IR

Indra Ribal, MSc

MSc
IHU Liryc, Université de Bordeaux, France
PG

Pauline Gut, MSc

MSc
University Hospital (CHUV) and University of Lausanne (UNIL), Switzerland
Valery Ozenne, PhD

Dr
CNRS, Aquitaine, France
GM

Géraldine Montier, MD

MD
Centre Hospitalier Universitaire de Bordeaux, France
JM

Jean-David Maes, MD

MD
Centre Hospitalier Universitaire de Bordeaux, France
TB

Thibault Boullé, MD

MD
Centre Hospitalier Universitaire de Bordeaux, France
GD

Guillaume Delclaux, MD

MD
Centre Hospitalier Universitaire de Bordeaux, France
AM

Aurélien Maillot, MSc

MSc
IHU Liryc, Université de Bordeaux, France
SS

Soumaya Sridi, MD

MD
Centre Hospitalier Universitaire de Bordeaux, France
BQ

Bruno Quesson, PhD

PhD
IHU Liryc, Université de Bordeaux, France
Matthias Stuber, PhD

Professor
University Hospital (CHUV) and University of Lausanne (UNIL)
Lausanne, Switzerland
HC

Hubert Cochet, MD, PhD

PROF/PhD
Centre Hospitalier Universitaire de Bordeaux
Pessac, Aquitaine, France
Aurélien Bustin, PhD

Junior Professor
IHU LIRYC
Bordeaux, Aquitaine, France

Disclosure(s):

Valery Ozenne, PhD: No relevant disclosure to display

Aurélien Bustin, PhD: No financial relationships to disclose

Background: Phase-sensitive inversion recovery is the reference imaging technique for the assessment of myocardial scars¹. Despite its ability to provide excellent contrast between healthy and scar tissue, small sub-endocardial scars can be challenging to detect due to poor scar-to-blood contrast. Joint bright- and black-blood late gadolinium enhancement (LGE) techniques² have been developed to provide images with improved scar contrast obtained from the black-blood images and detailed cardiac anatomy information obtained from the bright-blood images. Black-blood contrast is achieved after the manual selection by the radiographer of an optimal inversion time (TI) from a prior TI scout sequence. This often results in TI uncertainties and variability, and increased workload. In this work, we propose to automate the selection of the optimal TI by exploiting artificial intelligence, aiming to provide robust joint bright- and black-blood LGE cardiac imaging.

Methods: Algorithm: A TI scout is performed to collect single-shot black-blood images with varying TIs in the odd heartbeats and bright-blood images in the even heartbeat (Fig1A). The method to automate TI selection is organised in two steps (Fig1B). First, localization of the left ventricle is performed by automatically extracting the epicardium on the first collected bright-blood image using a U-Net architecture (trained with 186 patients). The centre of gravity of the resulting region-of-interest is computed to create a square mask centred on the resulting pixel (“focus box”), that is then propagated on all black-blood images.

In a second step, the sum of pixel intensities inside the focus box is computed for each image with inversion recovery. The lowest sum corresponds to the image with the darkest blood signal and corresponding TI is considered as the optimal value (Fig2A).

Experiments: Our proposed algorithm was validated on a dataset of 152 patients with known ischemic heart disease. Short-axis 2D images were collected on a 1.5-T system (Siemens, Aera) using the free-breathing SPOT sequence². TI values ranged from 60ms to 190ms with a 10ms increment. The algorithm accuracy was assessed by comparison with ground truth TI values obtained by an expert radiographer. Thirty focus box sizes, ranging from 2.3mm² to 20.3cm², were tested. Processing times were measured.

Results:

The proposed algorithm extracted the optimal TI value in 2.9s per patient (vs. ~25s for the expert). The algorithm reached its best accuracy with a focus box of size 5.1cm² (Fig2B) and predicted the correct TI in 151/152 patients (99.4%).
Examples of joint bright- and black-blood images collected using the TI generated by the algorithm are shown in Fig3 for three patients with myocardial infarction.

Conclusion:

The excellent accuracy obtained with the proposed fully automated algorithm suggests that uncertainties, variability, and workload inherent to more conventional manual TI selection approaches may be reduced with promise for future clinical implementation.