| Title |
Structural Segmentation for Sardiac Functional Analysis Based on Deep Learning |
| Authors |
이진아(Jin-A Lee) ; 곽근창(Keun-Chang Kwak) |
| DOI |
https://doi.org/10.5370/KIEEP.2023.72.2.120 |
| Keywords |
Deep Neural Network; Medical Image Processing; Cardiac; Image Segmentation |
| Abstract |
This paper is concerned with deep learning-based structural segmentation in order to perform functional analysis of the heart. The deep learning model used in this paper utilizes SegNet, MobileNetV2 and InceptionresnetV2 as the backbone networks within DeepLabV3+. The main objective is to segment the left ventricle in echocardiography images and segment the left ventricle, right ventricle, and myocardium in cardiac MRI images. The experimental results on the two benchmarking datasets confirm that the segmentation of the left ventricle is most accurate during the diastolic phase compared to the systolic phase. It was determined that the InceptionresnetV2 model is suitable for the CETUS MICCAI Challenge 2014 Dataset, while the MobileNetV2 model is appropriate for the ACDC MICCAI 2017 Dataset. |