Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox


MMFNet: A Multi-modality MRI Fusion Network for Segmentation of Nasopharyngeal Carcinoma

Jan 22, 2019
Huai Chen, Yuxiao Qi, Yong Yin, Tengxiang Li, Guanzhong Gong, Lisheng Wang


Share this with someone who'll enjoy it:


Segmentation of nasopharyngeal carcinoma (NPC) from Magnetic Resonance Images (MRI) is a crucial step in NPC radiotherapy. However, manually segmenting of NPC is a time-consuming and labor-intensive task. Additionally, single-modality MRI generally cannot provide enough information for the accurate delineation of NPC. Therefore, a multi-modality MRI fusion network (MMFNet) based on three modalities of MRI (T1, T2 and contrast-enhanced T1) is proposed to complete accurate segmentation of NPC. In the MMFNet, the backbone is designed as a multi-encoder-based network, consisting of several modality-specific encoders and one single decoder. It can be used to well learn both low-level and high-level features used implicitly for NPC segmentation in each modality of MRI. A fusion block is proposed in the MMFNet to effectively fuse low-level features from multi-modality MRI. It firstly recalibrates features captured from multi-modality MRI, which will highlight informative features and regions of interest. Then, a residual fusion block is utilized to fuse weighted features before merging them with features from decoder to keep balance between high-level and low-level features. Moreover, a training strategy named self-transfer is proposed to initialize encoders for multi-encoder-based network. It can stimulate encoders to make full mining of modality-specific MRI. The proposed method can effectively make use of information in multi-modality MRI. Its effectiveness and advantages are validated by many experiments and comparisons with the related methods.

* 37 pages, 11 figures 


   Access Paper Source



Share this with someone who'll enjoy it: