Detection of Fetal Cardiac Chamber Three Vessel Trachea View using Deep Learning

In fetal cardiology, echocardiographic imaging has been shown to aid in the diagnosis of fetal monitoring for the development of fetal heart function. The small heart size and the lack of expertise in fetal echocardiography of some sonography pose challenges in terms of research. The use of technology to enhance the image assists in the diagnosis of abnormalities important for optimal fetal cardiac assessment. Multi object detection is the most challenging but very informative task in basic machine learning. We must detect cardiac substructure precisely the condition of the fetal heart, especially in normal conditions. Knowing the normal heart substructure can provide information about the presence or absence of fetal heart abnormalities. Multiple ways to view fetal heart chambers, four chambers (4CH) basic standard cardiac scan, left and right ventricular outflow tract (LVOT/RVOT) focusing on structural pathways, and Three Tracheal Vessels (3VV/3VT) complex fetal heart scan confirms flow in the three blood vessels helps determine whether aortic arch problems are causing normal heart conditions or not. The ductus arteriosus (DUCT), ascending aorta (AoA), superior vena cava (SVC), and trachea. In particular, the framework performs detection using a deep learning detection model applied to 3VT ultrasound images. This study identifies the evaluation using the mean average precision (mAP). The results of the study detecting the normal condition of the fetal heart substructure gave good results on the results of the mAP evaluation of 87.55% and each object of DUCT, AoA, SVC, Trachea with AP values of 92%, 91%, 91%, 76%.

I. Introduction

A normal fetal heart is a very significant heart condition that must be known and analyzed because knowing normal heart conditions will reduce the risk of heart abnormalities. The heart has several different object structures to different viewing conditions. Find out the structure of the heart object, it can be seen from several view in the world of health, especially through ultrasound analysis. four chambers (4CH), left and right ventricular outflow tract (LVOT/RVOT), and Three Tracheal Vessels (3VV/3VT) is a view to see the heart condition on ultrasonography significantly [1] [2]. Each view has its own important role, such as the 4CH is the most important field that allows evaluation of the main cardiac structures, position, size, heart axis, contractility, and heart rhythm. RVOT and LVOT are views that have in common that only focus on the ventricles, the right ventricular outflow tract view for RVOT and left for LVOT are of the standard views in a fetal echocardiogram that only highlights the path from the ventricles to the pulmonary trunk [3]. 3VT is a view that tends to have a greater chance of detecting abnormalities in the fetal heart. In general, cardiac abnormalities can be seen based on size, spatial position, ductus arteriosus (DUCT), ascending aorta (AoA), superior vena cava (SVC), and trachea. Hence, this study proposes the process of detecting fetal heart chambers on video ultrasonography with a 3VT view based on its function and important role to see the occurrence of abnormalities in the fetal heart [4] [5]. This is an important role because variations between image data affect the process of heart growth and development so that a healthy heart condition can be seen from the intact heart chamber and according to its location. 3VT tends to have a greater chance of detecting abnormalities in the heart. In general, heart abnormalities can be seen based on the size, position, curvature, and holes that can provide in-depth information on the occurrence of abnormalities.