Chan-Ung Park

Computed tomography angiography (CTA)-based segmentation of the superficial temporal arteries (STAs) and facial vessels (FVs) is important for neurosurgical and reconstructive planning. Nevertheless, segmentation of STAs and FVs remains challenging because of their small caliber, tortuous courses, and proximity to high-intensity bony structures. This study aims to develop a deep learning framework for accurate automated segmentation of these craniofacial vessels. A single-input 3D nnU-Net v2 model was trained using raw CTA volumes, while a Fusion-based Vesselness Map (FVM) was constructed from multiscale vessel-enhancement filters to emphasize small vascular structures and suppress irrelevant regions such as the skull and skin. Instead of being used as an additional input channel, the FVM was incorporated into the loss function as a spatial prior to guide the network toward vessel boundaries and distal branches. In 72 clinical cases, the FVM-guided model improved segmentation accuracy compared with a baseline model trained with Dice Focal Loss, particularly in boundary delineation. For the STAs, the Average Symmetric Surface Distance decreased from 6.543 mm to 2.941 mm. Qualitative evaluation further showed reduced segmentation noise and fewer false positives near bone and distal branches. These findings suggest that integrating classical vessel enhancement into deep learning supervision can improve morphologically consistent craniofacial vessel segmentation and support preoperative surgical planning.

Introduction: Accurate identification of pelvic vascular structures is critical for safe and effective gynecologic cancer surgery, yet intraoperative recognition of the internal iliac artery, its branches, and corresponding venous pathways remains challenging due to complex anatomy and inter-individual variability. This study investigates the feasibility of fully automated, multi-class segmentation of both major pelvic vessels and smaller branches of the internal iliac artery and internal iliac vein using the nnU-Net v2 deep learning framework. Methods: Contrast-enhanced computed tomography angiography datasets from 47 patients were used for model training and evaluation. The nnU-Net v2 deep learning framework was employed to perform fully automated segmentation across nine vascular categories. These targets included major vessels, the internal iliac artery and vein, and an "other vessels" category consolidating surgically critical smaller branches, such as the obturator, uterine, and gluteal vessels. Model performance was quantitatively evaluated against ground truth annotations using the Dice Similarity Coefficient and Mean Surface Distance. Results: The model achieved high segmentation accuracy for large-diameter vessels such as the aorta and iliac arteries, with average Dice similarity coefficients exceeding 0.83 for arterial structures and 0.75 for venous structures. Smaller branches exhibited lower overlap scores but maintained anatomically coherent and clinically recognizable patterns, particularly for the obturator and gluteal arteries. Discussion: These findings demonstrate the potential of automated 3D vascular segmentation to enhance preoperative planning, anatomical education, and intraoperative safety in complex pelvic procedures. Furthermore, this represents the first systematic evaluation of the segmentation of major branches of the internal iliac artery in gynecologic oncology patients.