Breast MRI and Mammography Registration based on Breast Biomechanical FEM and Edge-Guided Unsupervised Registration...

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Radiology

Background:

• Breast cancer poses significant risks, necessitating accurate diagnostic tools.
• Mammography and breast MRI offer complementary information for diagnosis and treatment planning.
• Integrating 2D mammography and 3D MRI data is challenging due to dimensional differences.

Purpose of the Study:

• To develop a novel 2D/3D registration framework, BreastBioMorph (BBMorph), for aligning breast MRI and mammography images.
• To improve the accuracy of lesion localization and treatment planning by effectively mapping information between cross-modal breast images.

Main Methods:

• A biomechanical finite element model simulating breast compression was used for initial coarse alignment.
• An Edge-Guided unsupervised registration network refined deformation fields using breast edge constraints.
• The framework was validated using clinical breast data and compared against state-of-the-art methods.

Main Results:

• The BBMorph framework achieved a Dice coefficient of 89.05%, MI of 0.497, and SSIM of 0.815.
• Low deformation field distortion was observed (%|J_φ|≤0 of 0.104) with an average surface distance (ASSD) of 17.11.
• The method demonstrated superior performance compared to existing cross-modal registration approaches.

Conclusions:

• The proposed BBMorph framework provides effective 2D/3D registration for cross-modal breast images.
• This method offers a valuable tool for enhancing breast cancer diagnosis and treatment planning.
• Accurate image registration is crucial for combining complementary information from mammography and MRI.