Research On Medical Image Super Resolution Reconstruction Method Based On Residual Attention Network

Medical images are commonly used to determine the location,size,and shape of organs,and to determine the scope and physical properties of lesions,which are an important basis for intelligent medical diagnosis.Low-quality medical images have serious spots,noise,and weak boundaries between similar tissues,which affect the clarity of human organs and lesions in the image to a certain extent,and then affect the accuracy of target segmentation.This problem seriously hinders doctors’ subjective diagnosis and the accuracy of intelligent medical assisted diagnosis.Therefore,enhancing the internal texture details of medical images,strengthening tissue boundary information,and suppressing noise are of great significance for experts to diagnose diseases.This thesis takes lung CT images,brain MR images and transrectal ultrasound images of the prostate as the research objects,and proposes a medical image super-resolution reconstruction method AMSC-SR based on residual attention network.For single-scale and multi-scale reconstruction tasks,a single-scale AMSC-SR model and a multi-scale AMSC-SR model are established.For the single-scale AMSC-SR model,the residual network structure is first used to extract image features,then the channel attention module and spatial attention module are combined to enhance the effective features and suppress noise.At the same time,skip connection structure is introduced to enhance the transmission of shallow information and strengthen the texture and edge information of the medical image.Finally,the sub-pixel convolutional layer is used to achieve up-sampling of the feature layer,and complete the single-scale medical image reconstruction with a scale factor of ×2,×3or ×4.For the multi-scale AMSC-SR model,the single-scale AMSC-SR is used as the shared network of the model,and the network parameter sharing and sub-net parallel strategy are adopted to complete the multi-scale medical image reconstruction with scale factors of ×2,×3,and ×4 at one time.Experiments have confirmed that the proposed method can effectively reconstruct low-resolution medical images of different scales,accurately reconstruct the internal texture and edge information of medical images,and suppress noise effectively.The AMSC-SR method shows good reconstruction performance in terms of subjective qualitative and objective quantitative evaluation.