Magnetic Resonance Image Reconstruction Based on Multi-Scale and Attention Mechanisms

In order to improve the reconstruction quality of magnetic resonance imaging under undersampling, this paper proposes improvements based on the VS-Net network, which incorporates multi-scale residual convolution and attention mechanism modules into the network to achieve the reconstruction algorithm of magnetic resonance imaging under undersampling. This paper proposes a multi-scale feature MCRB module, which extracts and merges features from different scale feature maps to enhance the reconstruction of details and improve the resolution of reconstructed images. In addition, the CBAM attention mechanism module is incorporated into the network to amplify the detailed information and remove redundant features. The experimental results demonstrate that the proposed approach in this paper can reconstruct magnetic resonance images with higher quality and clearer texture details, indicating its high practical value.