The deep inferior epigastric artery perforator (DIEP) flap is the gold standard for autologous breast reconstruction following mastectomy, and involving transplantation of abdominal skin and fat while preserving muscle integrity. A critical yet challenging aspect of surgical planning is the accurate identification of perforator vessels - small arteries supplying blood to the transplanted tissue slab. Manual perforator selection from computed tomography angiography (CTA) scans is time-consuming, prone to error, and frequently resulting in intra-operative adjustments, prolonged surgery and increased risk of complications. We propose a fully-automated pre-operative planning framework designed to reliably map perforator vessels in CTA images. The system consists of a depth-aware annotation strategy that corrects spatial distortions associated with maximum intensity projection imaging; a patient-specific, anatomy-aware region of interest selection; and a deep-learning perforator segmentation pipeline leveraging a Swin UNETR architecture, trained with custom continuity-aware loss function and region-specific patch sampling. The method was validated on clinical CTA scans from 55 patients, using expert-guided annotations of the perforator vessels. The proposed method achieved a perforator segmentation median Dice similarity coefficient of 0.60 and a 95th-percentile Hausdorff distance of less than 18 mm, significantly improving vessel segmentation accuracy in fat regions over baseline Swin UNETR. These results demonstrate the clinical feasibility of automated segmentation, which has the potential to standardise surgical planning, enhance perforator selection precision, and ultimately improve patient outcomes in DIEP flap breast reconstruction.