GuidedMorph

Two-Stage Deformable Registration for Breast MRI

GuidedMorph is a two-stage deformable registration method specifically designed for breast MRI, addressing the challenge of aligning fibroglandular tissue (FGT) across different time points and imaging sessions.

Key Contributions

  • Two-stage architecture: Combines global alignment with local deformable registration for robust and accurate breast MRI registration
  • Superior performance: Outperforms existing deep learning-based registration models by over 13.01% in dense tissue Dice score on both internal and external datasets
  • Clinical relevance: Enables accurate breast density estimation and longitudinal analysis of breast tissue changes
  • Foundation model integration: Establishes benchmarks for foundation model-based registration in breast MRI

Applications

  • Breast density quantification and tracking over time
  • Longitudinal studies of breast tissue changes
  • Multi-temporal image analysis for cancer risk assessment
  • Clinical decision support for breast imaging
GuidedMorph architecture for breast MRI registration, achieving state-of-the-art performance in FGT alignment.

References