CMR-Analysis (including machine learning)

1354861 - Cardiac magnetic resonance with susceptibility artefacts: a novel artificial intelligence module for the assessment of left and right ventricular volumes and mass

Thursday, January 26, 2023

1:00 PM – 2:00 PM

Location: Pacific Jewel

Presenting Author(s)

MB

Mario Babbaro, MD

Cardiologist
IRCCS Policlinico San Donato, Italy

Primary Author(s)

mp

marco penso, BEng

Mr
Centro Cardiologico Monzino, Italy

Co-Author(s)

MB

Mario Babbaro, MD

Cardiologist
IRCCS Policlinico San Donato, Italy
SM

Sara Moccia, PhD

Phd candidate
Scuola Superiore Sant'Anna, Pisa, Italy
mG

marco Guglielmo, MD

Cardiologist
Utrecht University Medical Center, Netherlands
MP

Mauro Pepi, MD, PhD

Cardiologist
Centro Cardiologico Monzino, Italy
MR

Mark Rabbat, MD

Associat Professor
Loyola University Medical Center, United States
EC

Enrico G. Caiani, PhD, MS, BEng

Associate Professor of eHealth and Image processing
Politecnico di Milano
Milan, Lombardia, Italy
Gianluca Pontone, MD, PhD

Director of Peri Operative Cardiology and Cardiovascular Imaging Department
Centro Cardiologico Monzino
Milan, Lombardia, Italy

Background: Segmentation of cardiac magnetic resonance (CMR) images is an essential step for evaluating dimensional andfunctional ventricular parameters as ejection fraction (EF) but may be limited by artifacts, which represent the major challenge to automatically derive clinical information. The aim of this paper is to investigate the accuracy of a deep learning approach for automatic segmentation of cardiac structures from CMR images characterized by magnetic susceptibility artifact in patient with cardiac implantable electronic devices (CIED).

Methods: In this retrospective study, 230 patients (100 with CIED) who underwent clinically indicated CMR were used to developed and test a deep learning model. A novel convolutional neural network was proposed to extract the left (LV) and right (RV) ventricle endocardium and LV epicardium. In order to perform a successful segmentation, the proposed network takes advantage from a special attention module able to identify salient image regions even during local magnetic field inhomogeneities. Segmentation results were assessed against manual tracings and commercial CMR analysis software. An external dataset of 56 patients with CIED was used to assess model generalizability.

Results: In the internal datasets, on image with artifacts, the median Dice coefficient for end-diastolic LV cavity, LV myocardium and RV cavity, was 0.93, 0.77 and 0.87 and 0.91, 0.82, and 0.83 in end-systole, respectively. The proposed method reached higher segmentation accuracy than commercial software, with performance comparable to expert inter-observer variability (bias±95%LoA): LVEF 1±8% vs 3±9%, RVEF -2±15% vs 3±21%. In the external cohort, EF well correlated with manual tracing (intraclass correlation coefficient: LVEF 0.98, RVEF 0.93). The automatic approach was significant faster than manual segmentation in providing cardiac parameters (approximately 1.5 sec vs 450 sec).

Conclusion: Experimental results show that the proposed method reached promising performance in cardiac segmentation from CMR images with susceptibility artifacts and alleviates time consuming expert physician contour segmentation.