3D-NASE: A Novel 3D CT Nasal Attention-based Segmentation Ensemble

• DOI: 10.20944/preprints202503.0766.v1
• Publication Date: 2025-03-13T00:38:17Z
• Authors: Alessandro Pani; Luca Zedda; Davide Antonio Mura; Andrea Loddo; Cecilia Di Ruberto

Accurate segmentation of the nasal cavity and paranasal sinuses in CT scans is crucial for disease assessment, treatment planning, surgical navigation. It also facilitates advanced computational modeling airflow dynamics enhances endoscopic surgery preparation. This work presents a novel ensemble framework 3D that synergistically combines CNN-based transformer-based architectures, 3D-NASE. By integrating U-Net, UNETR, Swin UNETR with majority soft voting strategies, our approach leverages both local details global context to improve accuracy robustness. Results on NasalSeg dataset demonstrate proposed method surpasses previous state-of-the-art results by achieving 35.95% improvement DICE score reducing standard deviation 4.57%. These promising highlight potential advance clinical workflows diagnosis, navigation while promoting further research into computationally efficient highly accurate techniques.