Brain gray matter nuclei segmentation on quantitative susceptibility mapping using dual-branch convolutional neural network

作者：

Highlights：

• A 3D convolution neural network (CNN) method is proposed for gray matter nuclei segmentation on QSM and 3D T1WI images.

• A dual branch CNN structure is proposed to enlarge the fields of view of the network.

• Proposed method showed more prominent performance in nuclei segmentation than atlas-based and other deep-learning methods.

• The effectiveness on susceptibility value measurement is also discussed, where the proposed method presented better accuracy.

摘要

•A 3D convolution neural network (CNN) method is proposed for gray matter nuclei segmentation on QSM and 3D T1WI images.•A dual branch CNN structure is proposed to enlarge the fields of view of the network.•Proposed method showed more prominent performance in nuclei segmentation than atlas-based and other deep-learning methods.•The effectiveness on susceptibility value measurement is also discussed, where the proposed method presented better accuracy.

论文关键词：Convolutional neural network,Deep learning,Medical image segmentation,Gray matter nuclei,Quantitative susceptibility mapping

论文评审过程：Received 1 March 2021, Revised 27 December 2021, Accepted 5 February 2022, Available online 10 February 2022, Version of Record 25 February 2022.

论文官网地址：https://doi.org/10.1016/j.artmed.2022.102255