Deep morphological simplification network (MS-Net) for guided registration of brain magnetic resonance images

作者：

Highlights：

• Novel morphological simplification network (MS-Net) is proposed for simplifying deformable brain MR image registration.

• MS-Net is easily transferred across diverse image datasets.

• The intermediate images for pair-wise registration are derived from the original moving and fixed images, without the need of large population of subjects.

• Experiments show that the proposed method can achieve highly accurate registration performance, specifically on surface alignment over different datasets (i.e., NIREP, LPBA, IBSR, CUMC, and MGH).

摘要

•Novel morphological simplification network (MS-Net) is proposed for simplifying deformable brain MR image registration.•MS-Net is easily transferred across diverse image datasets.•The intermediate images for pair-wise registration are derived from the original moving and fixed images, without the need of large population of subjects.•Experiments show that the proposed method can achieve highly accurate registration performance, specifically on surface alignment over different datasets (i.e., NIREP, LPBA, IBSR, CUMC, and MGH).

论文关键词：Deformable image registration,Deep learning,Anatomical complexity

论文评审过程：Received 3 May 2019, Revised 27 November 2019, Accepted 15 December 2019, Available online 24 December 2019, Version of Record 5 January 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.107171