Semi-automated Processing of Real-Time CMR Scans for Left Ventricle Segmentation

• Published in: Biomedical Image Registration Vol. 10883; pp. 57 - 66
• Main Authors: Shahzad, Rahil, Fasshauer, Martin, Lelieveldt, Boudewijn P. F., Lotz, Joachim, van der Geest, Rob
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2018; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Left ventricle; Realtime-MR; Registration; Segmentation; Semi-automatic
• ISBN: 9783319922577; 3319922572
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-92258-4_6

We present a workflow for processing real-time cardiac MR (RT-CMR) scans for segmenting the left ventricle (LV) on short-axis slices (SAX). Our method is based on image registration, where the LV endocardium and epicardium are segmented by propagating a reference contour over all the frames of the RT-CMR SAX scans. Our method was evaluated on 19 subjects, the accuracy of the automatic LV endocardium and epicardium segmentation was compared to those defined manually. The proposed method obtained a dice similarity coefficient (DSC) of 0.94 and a mean surface-to-surface distance (MSD) measure of 0.89 ± 0.53 mm. Additionally, a number of automatically obtained clinical measures were compared to ground truth values. On average we obtained a Pearson’s correlation coefficient (R) of 0.94 (0.99–0.74).