BreastDeformNet

BreastDeformNet predicts non-rigid breast deformations between preoperative and intraoperative configurations, rapidly estimating patient-specific displacement vector fields (DVFs) from volumetric distance-field representations derived from MRI and surface data for image-guided breast surgery planning.

Key Features:

• Volumetric Distance-Field Representations: Utilizes volumetric distance-field representations derived from MRI and surface data for deformation estimation.
• Displacement Vector Fields (DVFs): Estimates patient-specific displacement vector fields (DVFs) that describe non-rigid tissue displacement.
• Anatomical Consistency: Produces anatomically consistent deformation estimates preserving anatomical relationships.
• Fast Estimation: Provides rapid inference of deformations suitable for intraoperative application.

Scientific Applications:

• Lesion localization for breast-conserving surgery: Addresses prone-to-supine MRI deformation to support preoperative localization of non-palpable early-stage breast tumors.
• Clinical evaluation: Validated on 10 clinical cases with an average tumor localization error of 15.46 ± 3.96 mm and an average tumor-skin projection distance of 13.38 ± 4.61 mm (DOI: 10.1109/ISBI56570.2024.10635174).

Methodology:

Constructs volumetric distance-field representations from MRI and surface data and employs a fully connected network trained with synthetic ground-truth displacement vector fields to predict patient-specific DVFs.