A CT-based predictive model for stent-induced vessel damage: application to type B aortic dissection

The distal stent-induced new entry (distal SINE) is a life-threatening device-related complication after thoracic endovascular aortic repair (TEVAR). However, risk factors for distal SINE are not fully determined, and prediction models are lacking. This study aimed to establish a predictive model for distal SINE based on the preoperative dataset.

Two hundred and six patients with Stanford type B aortic dissection (TBAD) that experienced TEVAR were involved in this study. Among them, thirty patients developed distal SINE. Pre-TEVAR morphological parameters were measured based on the CT-reconstructed configurations. Virtual post-TEVAR morphological and mechanical parameters were computed via the virtual stenting algorithm (VSA). Two predictive models (PM-1 and PM-2) were developed and presented as nomograms to help risk evaluation of distal SINE. The performance of the proposed predictive models was evaluated and internal validation was conducted.

Machine-selected variables for PM-1 included key pre-TEVAR parameters, and those for PM-2 included key virtual post-TEVAR parameters. Both models showed good calibration in both development and validation subsamples, while PM-2 outperformed PM-1. The discrimination of PM-2 was better than PM-1 in the development subsample, with an optimism-corrected area under the curve (AUC) of 0.95 and 0.77, respectively. Application of PM-2 in the validation subsample presented good discrimination with an AUC of 0.9727. The decision curve demonstrated that PM-2 was clinically useful.

This study proposed a predictive model for distal SINE incorporating the CT-based VSA. This predictive model could efficiently predict the risk of distal SINE and thus might contribute to personalized intervention planning.

This study established a predictive model to evaluate the risk of distal SINE based on the pre-stenting CT dataset and planned device information. With an accurate VSA tool, the predictive model could help to improve the safety of the endovascular repair procedure.

• Clinically useful prediction models for distal stent-induced new entry are still lacking, and the safety of the stent implantation is hard to guarantee. • Our proposed predictive tool based on a virtual stenting algorithm supports different stenting planning rehearsals and real-time risk evaluation, guiding clinicians to optimize the presurgical plan when necessary. • The established prediction model provides accurate risk evaluation for vessel damage, improving the safety of the intervention procedure.