Airway　segmentation　and　reconstruction　are　critical　for　preoperative　lesion　localization　and　surgical　planning　in　pulmonary　interventions.　However,　this　task　remains　challenging　due　to　the　intrinsically　complex　tree　structure　of　the　airway　and　the　imbalance　in　branch　sizes.　While　current　deep　learning　methods　focus　on　model　architecture　optimization,　they　underutilize　anatomical　priors　such　as　the　spatial　correlation　between　pulmonary　arteries　and　bronchi　beyond　geometric　grading　level　III.　To　address　this　limitation,　we　propose　a　dual-decoding　segmentation　network　(DDS-Net)　integrated　with　a　pulmonary-bronchial　extension　generative　adversarial　network　(PBE-GAN),　which　explicitly　embeds　artery-bronchus　adjacency　priors　to　enhance　distal　bronchial　identification.　Experimental　results　demonstrate　state-of-the-art　performance,　achieving　a　Dice　Similarity　Coefficient　(DSC)　of　88.46%,　Branch　Detection　Rate　(BD)　of　88.31%,　and　Tree　Length　Detection　Rate　(TD)　of　84.93%,　with　significant　improvements　in　detecting　peripheral　bronchi　near　pulmonary　arteries.　This　study　confirms　that　incorporating　anatomical　relationships　substantially　improves　segmentation　accuracy,　particularly　for　fine　structures.　Future　work　should　prioritize　clinical　validation　through　multi-center　trials　and　explore　integration　with　real-time　surgical　navigation　systems,　while　extending　similar　anatomical　synergy　principles　to　other　organ-specific　segmentation　tasks.　©　2025　Elsevier　Ltd