Fully Convolutional Neural Network for Cardiac Diagnosis Segmentation in Short-Axis Echocardiography

Abstract

This research discusses the application of Fully Convolutional Neural Networks (FCNs) for echocardiographic image segmentation as a diagnostic aid for heart conditions. Image segmentation is crucial in identifying cardiac structures, enabling accurate and timely diagnosis. In this study, the FCN-8 architecture is employed to perform segmentation on echocardiographic images with a short-axis view. The segmentation results are evaluated using several metrics, including Dice Coefficient, Intersection over Union (IoU), precision, and recall. Based on the evaluation, the model demonstrated good performance in distinguishing cardiac structures, with a Dice Coefficient of 0.85 and IoU of 0.79. This approach shows potential in assisting physicians in making faster and more accurate diagnoses for patients with heart disorders. This study concludes that FCN-8 can be an effective tool in supporting the diagnostic process in the medical field