← Back to Projects

Project

Shape description losses for medical image segmentation

Apr 1, 2021

Shape-aware objectives that inject global geometric and statistical structure into medical image segmentation training.

Shape-aware objectives that inject global geometric and statistical structure into medical image segmentation training.

  • Proposed several shape-based loss functions, including bounding-box-based objectives, to inject global geometric and statistical information into segmentation training.
  • Studied how physically meaningful shape descriptions improve organ segmentation across different anatomies and imaging modalities.
  • Segmentation
  • Shape Modeling
  • Loss Functions
  • Medical Imaging