Cross-modality Attention Method for Medical Image Enhancement

In this paper, we propose a cross-modality attention method (CMAM) for medical image enhancement especially in magnetic resonance imaging, which typically addresses the issue of exploiting the clean feature by generative model. To realize this goal, we distill the complementary information directly from different modalities of raw input images for reliable image generation. More specifically, our method integrates local features with exploiting the semantic high-order dependencies and thus explores attentional fields for robust feature representation learning. To evaluate the effectiveness of our CMAM, we conduct folds of experiments on the standard benchmark Brats 2019 dataset and experimental results demonstrate the effectiveness of CMAM.