<![CDATA[Congenital heart disease is a structural abnormality of the heart present from birth, affecting about 1% of all newborns, which make early detection of abnormal heart conditions is essential. Detection can be performed by calculating the traced area of end-systole and end-diastole segmentation in cardiac echocardiography videos. This study aims to perform segmentation on echocardiography images using the Dense-AIDAN method. The research workflow conducted in this study includes data collection and preparation, model development, and evaluation. The dataset used in this study is the public EchoNet-Dynamic echocardiography video dataset showing the four-chamber view of the heart. The echocardiography videos from the dataset are first converted into image frames. The image frames are generated based on the two tracings mentioned above. These images are then divided into training, validation, and test sets. The training images are used as input to train the Dense-AIDAN model. The trained model is then used to segment the left ventricle of the heart from the input test images. The implementation of the Dense-AIDAN method yields a Dice Similarity Coefficient (DSC) of 0.81 and an Intersection over Union (IoU) of 0.68. The study concludes that using DenseNet201 provides better segmentation results compared to ResNet50 on medical images, especially echocardiography images.]]>
Copyrights © 2026
