Path aggregation U-Net model for brain tumor segmentation

The deep neural network has been widely used in semantic segmentation, especially in tumor image segmentation. The segmentation performance of traditional methods cannot meet the high standard of clinical application. In this paper, we propose a new neural network model called path aggregation U-Net...

Full description

Saved in:
Bibliographic Details
Published in:Multimedia tools and applications 2021-06, Vol.80 (15), p.22951-22964
Main Authors: Lin, Fengming, Wu, Qiang, Liu, Ju, Wang, Dawei, Kong, Xiangmao
Format: Article
Language:English
Subjects:
Agglomeration
Artificial neural networks
Brain
Brain cancer
Coders
Computer Communication Networks
Computer Science
Data Structures and Information Theory
Image segmentation
Magnetic resonance imaging
Multimedia Information Systems
Neural networks
Special Purpose and Application-Based Systems
Tumors
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The deep neural network has been widely used in semantic segmentation, especially in tumor image segmentation. The segmentation performance of traditional methods cannot meet the high standard of clinical application. In this paper, we propose a new neural network model called path aggregation U-Net (PAU-Net) model for brain tumor segmentation with multi-modality magnetic resonance imaging (MRI). Specifically, we shorten the distance between output layers and deep features by bottom-up path aggregation encoder (PA), reducing the introduction of noises. We present the enhanced decoder (ED) to reserve more intact information. The efficient feature pyramid (EFP) is used to improve mask prediction further, using fewer resources to complete the feature pyramid effect. Finally, experiments in BraTS2017 and BraTS2018 datasets are performed. The results show that the proposed method outperforms state-of-the-art methods.
ISSN:1380-7501
1573-7721
DOI:10.1007/s11042-020-08795-9