Automated Segmentation of the Deep Inferior Epigastric Artery on CTA Using Deep Learning for Preoperative Planning of DIEP Flap Breast Reconstruction

Accurate segmentation of the deep inferior epigastric artery (DIEA) on computed tomography angiography (CTA) is important for preoperative planning of deep inferior epigastric perforator (DIEP) flap breast reconstruction. This retrospective study established a dedicated DIEA dataset and developed a deep learning model for automated segmentation. A total of 4446 high-resolution abdominal CTA images from 31 patients were curated and annotated at the pixel level using a multi-reviewer protocol. An improved ResU-Net++ model was trained with Tversky loss, AdamW optimization, and cosine annealing, and was compared with U-Net, ResU-Net, and ResU-Net++ using Accuracy, Sensitivity, Intersection over Union, Dice Similarity Coefficient (DSC), and Hausdorff Distance (HD). The proposed model achieved a Dice Similarity Coefficient (DSC) of 0.6807, an IoU of 0.5556, and a Sensitivity of 0.7326, representing the highest values for these metrics among the evaluated models. The complete automated processing pipeline required 0.0093±0.0015 s per image, compared with 19.85±5.75 s for manual segmentation (p<0.001). These findings support the technical feasibility of automated DIEA segmentation on abdominal CTA. Its potential application in DIEP flap preoperative planning and clinical workflows requires further prospective validation.