An Unsupervised GAN-based Quality-enhanced Medical Image Fusion Network

Medical image fusion technology can improve the precision of clinical diagnosis by fusing medical information from different modalities. However, the quality of fusion is restricted due to the particular imaging mechanism. This paper proposes a quality-enhanced medical image fusion algorithm based on a generative adversarial network for the lossless fusion of MRI and PET images. It consists of a lightweight image enhancement depth network to make the quality of the fused image suit human vision perceptual system better and a generative adversarial network to enhance texture details and edge information further. Our model is unsupervised and does not require paired fused images for training. The test results show that our algorithm performs better in both subjective visual effects and objective evaluation metrics.