Medical image super-resolution with deep networks

Can Zhao, Samuel W. Remedios, Shuo Han, Bowen Li, Jerry L. Prince

Research output: Chapter in Book/Report/Conference proceedingChapter


This chapter explores super-resolution (SR) algorithms with deep networks and their applications in medical imaging. SR algorithms aim to improve image quality by enhancing image resolution so that more details are apparent. In medical imaging modalities such as computed tomography and magnetic resonance imaging, SR algorithms can improve image resolution without requiring changes to the imaging protocols. Being able to discern regions of interest better, physicians can potentially provide more reliable diagnoses. Similarly, many downstream image processing tasks such as registration and segmentation can benefit from increased resolution and thus provide more accurate results. In recent years, the state-of-the-art SR methods have advanced rapidly due to the advent of deep neural networks. In this chapter, we first introduce basic concepts and a brief history of SR methods. We then summarize SR with deep networks in medical imaging by focusing on three aspects: data schemes, network architectures, and loss functions. Finally, we examine real-world applications of SR in different medical imaging modalities and tasks. Through the use of deep networks, SR is likely to become routine in medical imaging applications in the future.

Original languageEnglish (US)
Title of host publicationBiomedical Image Synthesis and Simulation
Subtitle of host publicationMethods and Applications
Number of pages21
ISBN (Electronic)9780128243497
ISBN (Print)9780128243503
StatePublished - Jan 1 2022


  • Deep learning
  • Image quality
  • Super-resolution

ASJC Scopus subject areas

  • Computer Science(all)


Dive into the research topics of 'Medical image super-resolution with deep networks'. Together they form a unique fingerprint.

Cite this