Region of interest optimization for radiation therapy of breast cancer

The goal of this study was to investigate how the choice of the region of interest (ROI) affects the registration results of surface imaging for daily positioning of breast cancer patients.

The AlignRT system (VisionRT, London) and the XVI Cone beam CT (CBCT; Elekta, Stockholm) installed on two Versa HD linacs (Elekta) were used in this study, which included 28 patients (160 fractions). In the clinical workflow, patients were prepositioned with AlignRT and then shifted in 6 degrees of freedom (DOF) according to the CBCT. A new reference capture was taken immediately afterward. Retrospectively, the surface capture resulting from prepositioning was registered to the latest reference capture. By varying the ROI used for registration, the surface-based results were optimized in terms of minimizing the deviation to the clinically applied CBCT shifts. Two sets of ROIs were used: one obtained by applying a variable margin to the breast surface, another by combining ROIs of anatomical structures, including the sternum and contralateral breast.

Registration results showed significant differences from one ROI to another. Generally, the results improved with increasing ROI size, especially for rotational DOFs. ROIs, including the axilla or supraclavicular lymph drainage region, did not yield an improved registration result. On the other hand, an ROI comprising the breast surface, sternum, and a belt caudal to the breasts decreased the average magnitude of the translational and rotational deviations by 6.6% and 30.8% (p < 0.01), respectively, compared to the breast surface only results.

The influence of the ROI choice on surface imaging registration results was analyzed and the surface-based shifts were compared to clinically applied CBCT shifts. An optimal ROI for the treatment of breast cancer patients, consisting of the breast surface, sternum, and a belt, was identified.