Deep learning for medical image segmentation

Yading Yuan, Ronald Levitin, Zaid Siddiqui, Richard Bakst, Michael Buckstein, Evan Porter

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review


Accurate and reliable automated segmentation plays a vital role in improving consistency, efficiency and quality of patient care in clinical radiation therapy process, while also enabling comprehensive quantitative image analysis for assessing treatment outcomes on a large scale. In recent years, deep learning-based methods, which seamlessly integrate information ranging from global semantic context to intricate details within a unified end-to-end framework, have demonstrated substantially superior performance than traditional algorithms in numerous tasks involving tumor and/or organ segmentation. In this chapter, we firstly present the rationale of using deep learning for medical image segmentation, then we discuss several practical considerations when developing a deep learning model for a particular segmentation task, including image pre-processing, image patch selection, data augmentation, model fusion and output uncertainty assessment. Finally, we express our perspectives on the significance of international image analysis competitions in introducing innovative ideas and models, as well as in educating emerging researchers in the field of auto-segmentation in radiotherapy.

Original languageEnglish
Title of host publicationMachine Learning and Artificial Intelligence in Radiation Oncology
Subtitle of host publicationA Guide for Clinicians
Number of pages29
ISBN (Electronic)9780128220009
StatePublished - 1 Jan 2023


  • Convolutional neural networks
  • Deep learning
  • Image segmentation
  • Machine learning
  • Medical imaging


Dive into the research topics of 'Deep learning for medical image segmentation'. Together they form a unique fingerprint.

Cite this