Deep Learning for the Detection of Frames of Interest in Fetal Heart Assessment from First Trimester Ultrasound

• Published in: Advances in Computational Intelligence Vol. 12861; pp. 3 - 14
• Main Authors: Stoean, Ruxandra, Iliescu, Dominic, Stoean, Catalin, Ilie, Vlad, Patru, Ciprian, Hotoleanu, Mircea, Nagy, Rodica, Ruican, Dan, Trocan, Rares, Marcu, Andreea, Atencia, Miguel, Joya, Gonzalo
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2021; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Classification; Deep learning; Fetal heart; First trimester pregnancy; Ultrasound
• ISBN: 9783030850296; 3030850293
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-85030-2_1

The current paper challenges convolutional neural networks to address the computationally undebated task of recognizing four key views in first trimester fetal heart scanning (the aorta, the arches, the atrioventricular flows and the crossing of the great vessels). This is the primary inspection of the heart of a future baby and an early recognition of possible problems is important for timely intervention. Frames depicting the views of interest were labeled by obstetricians and given to several deep learning architectures as a classification task against other irrelevant scan sights. A test accuracy of 95% with an F1-score ranging from 90.91% to 99.58% for the four key perspectives shows the potential in supporting heart scans even from such an early fetal age, when the heart is still quite underdeveloped .