Multiple improved residual networks for medical image super-resolution

Abstract The rapid development of deep learning has resulted in great breakthroughs in image super-resolution reconstruction technology in medical imaging modalities. The application of artificial intelligence to medical image processing has been the focus of scholars both domestically and internationally in recent years. Due to the fast super-resolution convolutional neural network (FSRCNN) algorithm has fewer convolutional layers and lacks the correlation between the feature information of adjacent convolutional layers, it is difficult to be used to extract deep information of an image, and the super-resolution rate of the image reconstruction effect is not good. To solve this problem, we propose the multiple improved residual network (MIRN) super-resolution reconstruction method. First, MIRN designs the residual blocks connected by multi-level skips to build multiple improved residual block (MIRB) modules. A deep residual network with multi-level skip connection is used to solve the lack of correlation between the characteristic information of adjacent convolutional layers. Then, the stochastic gradient descent method (SGD) is used to train a deep residual network connected by multi-level jumpers with an adjustable learning rate strategy to obtain a super-resolution reconstruction model of the network. Finally, the low-resolution image is input in the MIRN super-resolution reconstruction model, and the residual block obtains the predicted residual eigenvalues and then combines the residual image and the low-resolution image into a high-resolution image. Most quantitative and qualitative evaluations on benchmark datasets demonstrate that the proposed model can better reconstruct the details and textures of images and avoid the over-smoothing of medical images after iteration, and the performance of the proposed algorithm is revealed to be better than that of existing state-of-the-art methods.