PET and MRI Medical Image Fusion Based on Densely Connected Convolutional Networks

Medical image fusion technology has played an increasingly important role in many clinical applications by extracting complementary information from medical images of different morphologies and solving the shortcomings that exist in some medical images. In this paper, we use a densely connected convolutional networks to fuse the multimodal medical image positron emission tomography(PET) and the magnetic resonance imaging(MRI), but compared with the traditional convolutional neural network, it is different; The model consists of convolutional layer and dense block, the output of the previous layer is in turn used as the input of the next layer, the encoder obtains valuable information from the source image, the fusion layer fuses these features, and finally the decoder reconstructs the combined image, thus completing the fusion of the picture. The images obtained by fusion are then analyzed quantitatively and qualitatively, and the results obtained by our method used are compared with those obtained by five other image fusion algorithms. The qualitative analysis shows that our method achieves better results in terms of visual quality and retains more texture information and image features; the quantitative analysis of the comparison data shows that our method achieves a maximum in two indicators that the structure is more similar to that of the source images.