Supine-Positioned Breast Deformations

In 2021, we measured breast deformation in the supine position using our image guidance system in n=6 healthy volunteers. We evaluated how well rigid registration compensated for these breast deformations from supine imaging to surgical position. While rigid registration frameworks have been shown to reduce reoperation rates in breast conserving surgery (BCS) [Barth et. al., 2019], nonrigid deformation will be an essential consideration in continuing to improve BCS outcomes.

Main contributions of this work:

  • There are large nonrigid shape changes that occur between supine imaging and surgical presentation
  • Simple abduction of the arm results in significant shifts and shape changes in breast tissue (28.9 mm on average)
  • Residual error after image-to-physical rigid alignment can be large - on the order of 10–30 mm on the breast surface.
  • Even with no change in arm pose, simple intraprocedural changes such as small rotations of the operating room table can result in significant alignment errors after rigid alignment (image-to-physical rigid registration). On the surface, mean fiducial registration error ~ 6 mm, maximum target registration error ~12.5 mm on average.
  • For 90 degree abduction of the arm, 50% of deformation is in the inferior-superior direction, and the remaining directionality is approximately evenly split between the medial-lateral and anterior-posterior directions.

View PDF

Breast Deformation from Supine Imaging to Surgery within an Image Guided Approach from VISE on VISE's Vimeo page.

Patient Specific Breast Modeling

Recently, we modified the Linearized Iterative Boundary Reconstruction method, established by Heiselman et. al., for breast specific inputs. The nonrigid registration is driven with sparse data compatible with acquisition during breast surgery, including corresponding surface fiducials in image and physical space, sparse chest wall contours, and the intra-fiducial skin surface. The data suggest that a computational imaging approach can account for shape change to enhance surgical guidance during BCS.

View PDF

For more relevant modeling, check out how Dr. Heiselman's Linearized Iterative Boundary Reconstruction method was expanded to model pressure changes during image guided liver surgery, and to incorporate alternative basis functions (Kelvinlets) originally designed for deformable animations at Pixar!