Deep Learning for Medical Image Super-Resolution: A Review

Medical image super-resolution (SR) reconstruction technology has a wide range of scenarios and important values in medical diagnosis. The use of SR reconstruction algorithms can enhance low-resolution (LR) medical images into high-resolution (HR) medical images, thus overcoming the problems of low spatial resolution caused by the limitations of physical equipment conditions of image acquisition and the own limitations of imaging principles, and can present the details of human organs or tissues more clearly, especially making the local feature information of lesion points, thus assisting doctors to make a more accurate diagnosis in the clinic. This paper reviews the research on medical image SR reconstruction methods at home and abroad, firstly introduces the performance indexes for evaluating image SR reconstruction methods, then systematically describes the medical image SR reconstruction technology based on deep learning methods, and finally analyzes the problems that still exist in this field and points out the future research directions.