Automated volumetric surveillance of Type B aortic dissection (TBAD) is limited by manual segmentation overhead (60-75 min) in computed tomographic angiography (CTA). We propose AORTA-SCAN, a hierarchical deep-learning network designed to identify the true and false lumina of the aortic dissection (AD), in both primary and post-TEVAR CTA. Along with the traditional 1D biomarker (1D (aortic diameter), the framework provides the perfusion dissection index (PDI), which is a 3D clinical indicator of aortic remodelling. The proposed model shows a significant improvement in false lumen dice (0.84 vs 0.71, p < 0.001), with respect to state-of-the-art models, and high clinical agreement for the PDI (mean bias -0.17%) compared to the clinical reference. The pipeline reduces processing time to about 10 minutes, supporting scalable 3D volumetric surveillance.