Room: AAPM ePoster Library
Purpose: Deformable image registration (DIR) can add value in accurate delineation of tumor bed for breast conserving radiotherapy. However, this can be challenging due to large deformation and non-correspondence caused by tumor resection and clip insertion. In this study, a new B-spline based DIR method was developed and tested on clinical patient data.
Methods: The traditional B-spline based DIR method was modified with introduction of point metric and mask. The point metric is used to regularize an intensity-based B-spline registration which minimizes the distance of two point sets with known correspondence. And the mask was used for reducing the influence of non-corresponding regions. One diagnostical CT was acquired one week before breast conserving surgery and another planning CT was acquired during CT simulation for radiotherapy. Total 26 image sets of 13 patients were collected retrospectively between November 2017 and May 2019. The performances of the methods were evaluated based on target registration error (TRE), Jacobian determinant and visual assessment.
Results: The method with both point metric and mask has the best performance and the method with only point metric is secondary. The method with only mask shows less improvement than the method with traditional B-spline based DIR. The TRE of the method with both point metric and mask is 2.39±0.56mm, 3.61±3.92mm, 3.50±1.88mm for soft tissue, rigid structure, boundary respectively. And the TRE with traditional B-spline based DIR is 6.86±4.48mm, 3.38±3.55mm, 5.65±5.22mm for soft tissue, rigid structure, boundary respectively. The Jacobian determinant of the deformation field was > 0 within all the patients. Visual assessment is much better in the method with point metric.
Conclusion: The method with both point metric and mask outperforms the method with traditional B-spline based DIR, and has great potential in accurate delineation of tumor bed for breast conserving radiotherapy.