Computed Tomography-Based Periaortic Adipose Tissue Imaging Phenotypes in Type B Aortic Dissection: A Retrospective...

The periaortic microenvironment visible on routine computed tomography angiography remains incompletely characterized in type B aortic dissection (TBAD). We evaluated computed tomography angiography-based periaortic adipose tissue (PAAT) phenotypes associated with prevalent TBAD and explored a prespecified Tear Risk Index for post-thoracic endovascular aortic repair aortic dilation.

This retrospective case-control study included 100 patients with TBAD undergoing primary thoracic endovascular aortic repair and 50 normal controls frequency matched by age and sex. Deep learning-assisted, radiologist-reviewed segmentation quantified multilayer PAAT and aortic morphology. Interreader reproducibility was assessed using intraclass correlation coefficients and Bland-Altman analyses. Group comparisons, receiver operating characteristic analyses, and adjusted logistic regression were performed.

Compared with controls, patients with TBAD had larger aortic volume and higher 3-mm PAAT volume, fat attenuation index, and fragmentation index. Aortic volume showed an area under the curve of 0.966 (95% CI, 0.941-0.991); corresponding values for 3-mm PAAT volume, fat attenuation index, and fragmentation index were 0.614, 0.819, and 0.716, respectively. Interreader intraclass correlation coefficients for selected 3-mm PAAT measures ranged from 0.822 to 0.907. Principal TBAD-control PAAT findings remained supported after false discovery rate correction, whereas post-thoracic endovascular aortic repair PAAT differences were not robust. Tear Risk Index showed limited exploratory discrimination for dilation status.

Selected 3-mm PAAT measures characterize a reproducible computed tomography angiography-based imaging phenotype associated with prevalent TBAD and may complement conventional aortic morphology. These findings are hypothesis generating, and Tear Risk Index should not be used as a standalone clinical predictor.