Single MR image super-resolution via channel splitting and serial fusion network

作者：

Highlights：

• A novel Serial Local Feature Fusion strategy is presented for channel splitting.

• The dilemma between model trainability and scale is eased for MR image SR.

• We have experimentally verified the overfitting problem of MR training samples.

• Aggressive channel splitting will exacerbate the problem of overfitting.

摘要

•A novel Serial Local Feature Fusion strategy is presented for channel splitting.•The dilemma between model trainability and scale is eased for MR image SR.•We have experimentally verified the overfitting problem of MR training samples.•Aggressive channel splitting will exacerbate the problem of overfitting.

论文关键词：Convolutional neural network,Magnetic resonance imaging,Channel splitting,Super-resolution,Serial fusion

论文评审过程：Received 19 October 2021, Revised 23 March 2022, Accepted 24 March 2022, Available online 31 March 2022, Version of Record 22 April 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.108669