TY - GEN
T1 - A Two-Stage U-Net Model for 3D Multi-class Segmentation on Full-Resolution Cardiac Data
AU - Wang, Chengjia
AU - MacGillivray, Tom
AU - Macnaught, Gillian
AU - Yang, Guang
AU - Newby, David
N1 - Funding Information:
This work is funded by BHF Centre of Cardiovascular Science and MICCAI 2017 Multi-Modality Whole Heart Segmentation (MM-WHS) challeng.
Publisher Copyright:
© 2019, Springer Nature Switzerland AG.
PY - 2019/2/14
Y1 - 2019/2/14
N2 - Deep convolutional neural networks (CNNs) have achieved state-of-the-art performances for multi-class segmentation of medical images. However, a common problem when dealing with large, high resolution 3D data is that the volumes input into the deep CNNs has to be either cropped or downsampled due to limited memory capacity of computing devices. These operations can lead to loss of resolution and class imbalance in the input data batches, thus downgrade the performances of segmentation algorithms. Inspired by the architecture of image super-resolution CNN (SRCNN), we propose a two-stage modified U-Net framework that simultaneously learns to detect a ROI within the full volume and to classify voxels without losing the original resolution. Experiments on a variety of multi-modal 3D cardiac images have demonstrated that this framework shows better segmentation performances than state-of-the-art Deep CNNs with trained with the same similarity metrics.
AB - Deep convolutional neural networks (CNNs) have achieved state-of-the-art performances for multi-class segmentation of medical images. However, a common problem when dealing with large, high resolution 3D data is that the volumes input into the deep CNNs has to be either cropped or downsampled due to limited memory capacity of computing devices. These operations can lead to loss of resolution and class imbalance in the input data batches, thus downgrade the performances of segmentation algorithms. Inspired by the architecture of image super-resolution CNN (SRCNN), we propose a two-stage modified U-Net framework that simultaneously learns to detect a ROI within the full volume and to classify voxels without losing the original resolution. Experiments on a variety of multi-modal 3D cardiac images have demonstrated that this framework shows better segmentation performances than state-of-the-art Deep CNNs with trained with the same similarity metrics.
KW - Cardiac CT/MR
KW - Convolutional neural networks
KW - High resolution
KW - Image segmentation
UR - http://www.scopus.com/inward/record.url?scp=85064048425&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-12029-0_21
DO - 10.1007/978-3-030-12029-0_21
M3 - Conference contribution
AN - SCOPUS:85064048425
SN - 9783030120283
T3 - Lecture Notes in Computer Science
SP - 191
EP - 199
BT - Statistical Atlases and Computational Models of the Heart. Atrial Segmentation and LV Quantification Challenges. STACOM 2018
PB - Springer
T2 - 9th International Workshop on Statistical Atlases and Computational Models of the Heart: Atrial Segmentation and LV Quantification Challenges
Y2 - 16 September 2018 through 16 September 2018
ER -