Next Article in Journal
Turbine Blade Three-Wavelength Radiation Temperature Measurement Method Based on Reflection Error Correction
Next Article in Special Issue
Segmentation of Liver Anatomy by Combining 3D U-Net Approaches
Previous Article in Journal
Influence of the Periapical Status of the Posterior Maxillary Teeth on the Width of the Schneiderian Membrane of the Maxillary Sinus Mucosa
Previous Article in Special Issue
Virtual UV Fluorescence Microscopy from Hematoxylin and Eosin Staining of Liver Images Using Deep Learning Convolutional Neural Network
Article

Whole Heart Segmentation Using 3D FM-Pre-ResNet Encoder–Decoder Based Architecture with Variational Autoencoder Regularization

Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 31000 Osijek, Croatia
*
Author to whom correspondence should be addressed.
These authors contributed equally to this work.
Academic Editor: Anna Fabijańska
Appl. Sci. 2021, 11(9), 3912; https://doi.org/10.3390/app11093912
Received: 30 March 2021 / Revised: 20 April 2021 / Accepted: 21 April 2021 / Published: 26 April 2021
(This article belongs to the Special Issue Artificial Intelligence for Medical Image Analysis)
An accurate whole heart segmentation (WHS) on medical images, including computed tomography (CT) and magnetic resonance (MR) images, plays a crucial role in many clinical applications, such as cardiovascular disease diagnosis, pre-surgical planning, and intraoperative treatment. Manual whole-heart segmentation is a time-consuming process, prone to subjectivity and error. Therefore, there is a need to develop a quick, automatic, and accurate whole heart segmentation systems. Nowadays, convolutional neural networks (CNNs) emerged as a robust approach for medical image segmentation. In this paper, we first introduce a novel connectivity structure of residual unit that we refer to as a feature merge residual unit (FM-Pre-ResNet). The proposed connectivity allows the creation of distinctly deep models without an increase in the number of parameters compared to the pre-activation residual units. Second, we propose a three-dimensional (3D) encoder–decoder based architecture that successfully incorporates FM-Pre-ResNet units and variational autoencoder (VAE). In an encoding stage, FM-Pre-ResNet units are used for learning a low-dimensional representation of the input. After that, the variational autoencoder (VAE) reconstructs the input image from the low-dimensional latent space to provide a strong regularization of all model weights, simultaneously preventing overfitting on the training data. Finally, the decoding stage creates the final whole heart segmentation. We evaluate our method on the 40 test subjects of the MICCAI Multi-Modality Whole Heart Segmentation (MM-WHS) Challenge. The average dice values of whole heart segmentation are 90.39% (CT images) and 89.50% (MRI images), which are both highly comparable to the state-of-the-art. View Full-Text
Keywords: artificial intelligence; cardiac CT; cardiac MRI; deep learning; ResNet; variational autoencoder; whole heart segmentation artificial intelligence; cardiac CT; cardiac MRI; deep learning; ResNet; variational autoencoder; whole heart segmentation
Show Figures

Figure 1

MDPI and ACS Style

Habijan, M.; Galić, I.; Leventić, H.; Romić, K. Whole Heart Segmentation Using 3D FM-Pre-ResNet Encoder–Decoder Based Architecture with Variational Autoencoder Regularization. Appl. Sci. 2021, 11, 3912. https://doi.org/10.3390/app11093912

AMA Style

Habijan M, Galić I, Leventić H, Romić K. Whole Heart Segmentation Using 3D FM-Pre-ResNet Encoder–Decoder Based Architecture with Variational Autoencoder Regularization. Applied Sciences. 2021; 11(9):3912. https://doi.org/10.3390/app11093912

Chicago/Turabian Style

Habijan, Marija, Irena Galić, Hrvoje Leventić, and Krešimir Romić. 2021. "Whole Heart Segmentation Using 3D FM-Pre-ResNet Encoder–Decoder Based Architecture with Variational Autoencoder Regularization" Applied Sciences 11, no. 9: 3912. https://doi.org/10.3390/app11093912

Find Other Styles
Note that from the first issue of 2016, MDPI journals use article numbers instead of page numbers. See further details here.

Article Access Map by Country/Region

1
Back to TopTop