A coherent point drift algorithm for breast image registration via surface markers

Abstract

Breast Magnetic Resonance Imaging (MRI) is a reliable imaging tool for localization and evaluation of lesions prior to breast conserving surgery (BCS). MR images typically will be used to determine the size and location of the tumours before making the incision in order to minimize the amount of tissue excised. The arm position and configuration of the breast during and prior to surgery are different and one question is whether it would be possible to match the two configurations. This matching process can potentially be used in development of tools to guide surgeons in the incision process. Recently, a Thin- Plate-Spline (TPS) algorithm has been proposed to assess the feasibility of breast tissue matching using fiducial surface markers in two different arm positions. The registration algorithm uses the surface markers only and does not employ the image intensities. In this Thesis, I apply and evaluate a coherent point drift (CPD) algorithm for registration of 3D breast MR images of six patient volunteers. In particular, we evaluate the results of the previous TPS registration technique to the proposed rigid, affine, and deformable CPD registration algorithms on the same patient datasets. The results suggest that the CPD deformable registration algorithm is superior in correcting the motion of the breast compared to CPD rigid, affine and TPS registration algorithms. The CPD registration results reported in this thesis took 0.2 to 0.4 seconds, which is significantly lower than the computation time using TPS (under a minute).