Application of delayed-enhancement MRI in lumpectomy cavity delineation for patients after breast-conserving surgery in prone radiotherapy using deformable image registration

Objective To assess the feasibility of delayed-enhancement MRI in contouring the lumpectomy cavity (LC) for patients with invisible seroma or a low cavity visualization score (CVS≤2) in the excision cavity after breast-conserving surgery (BCS). Methods Twenty-six patients with stageT1-2N0M0 who underwent prone radiotherapy after BCS were recruited. The LC delineated on CT simulation images was denoted as LCCT. The LCs delineated on T2WI, as well as on different delayed phases (2-, 5-and 10-minute) of delayed-enhancement T1WI were defined as LCT2, LC2T1, LC5T1 and LC10T1, respectively. Subsequently, the volumes and locations of the LCs were compared between CT simulation images and different sequences of MR simulation images using deformable image registration. Results The volumes of LCT2, LC2T1, LC5T1 and LC10T1 were all larger than that of LCCT. A statistical significance was found between the volume of LCCT and those of LC2T1 or LC5T1, respectively (both P<0.05). The conformal index (CI), degree of inclusion (DI), dice similarity coefficient (DSC) and the distance between the center of mass of the targets (COM) of LCCT-LC10T1 were better than those of LCCT-LCT2, LCCT-LC2T1 and LCCT-LC5T1, however, there was no statistical difference among them (all P>0.05). Conclusions It is feasible to delineate the LC based on prone delayed-enhancement MR simulation images in patients with low CVS after BCS. Meanwhile, the LCs derived from prone delayed-enhancement T1WI of 10-minute are the most similar with those derived from prone CT simulation scans using titanium clips, regardless of the volumes and locations of LCs.