Random effects during training: Implications for deep learning-based medical image segmentation.

Journal: Computers in biology and medicine
PMID: 39096609

Abstract

BACKGROUND: A single learning algorithm can produce deep learning-based image segmentation models that vary in performance purely due to random effects during training. This study assessed the effect of these random performance fluctuations on the reliability of standard methods of comparing segmentation models.

Authors

  • Julius Åkesson
    Clinical Physiology, Department of Clinical Sciences Lund, Lund University, Skåne University Hospital, Lund, Sweden. julius.akesson@med.lu.se.
  • Johannes Töger
    Clinical Physiology, Department of Clinical Sciences Lund, Lund University, Lund, Sweden; Department of Biomedical Engineering, Faculty of Engineering, Lund University, Lund, Sweden.
  • Einar Heiberg
    Clinical Physiology, Department of Clinical Sciences Lund, Lund University, Skane University Hospital, Lund, Sweden.

Keywords

External Resources

Popular Topics
Recent Journals