Circulating Metalloproteinase Inhibitors and Artificial Intelligence-Derived Radiomics Predict Persistent Endoleak After Endovascular Abdominal Aortic Aneurysm Repair: A Multimodal Biological-Imaging Model.

Authors

Abstract

Persistent endoleak after endovascular abdominal aortic aneurysm repair (EVAR) reflects ongoing aneurysm sac biological activity and remains insufficiently predicted by anatomical surveillance alone. We aimed to evaluate whether integration of circulating matrix metalloproteinases (MMPs), tissue inhibitors of metalloproteinases (TIMPs), and artificial intelligence (AI)-derived radiomic features from computed tomographic angiography (CTA) improves prediction of persistent endoleak. We conducted a retrospective observational study including 62 consecutive patients undergoing elective infrarenal EVAR between January 2020 and December 2024. Plasma levels of MMP-1, MMP-2, MMP-9, MMP-10, TIMP-1, and TIMP-2 were measured preoperatively and at 1 and 12 months post-EVAR. Computed tomography angiography data sets were analyzed using a standardized semi-automated radiomics pipeline with reproducibility testing and penalized feature selection. The primary endpoint was persistent endoleak at 12 months. Multivariable logistic regression was constructed with event-per-variable control and internally validated using bootstrap resampling. Persistent endoleak occurred in 18 patients (29%). Patients with persistent endoleak exhibited significantly higher MMP-9 levels, lower TIMP-1 and TIMP-2 levels, and markedly elevated MMP-9/TIMP-1 ratios at 1 month (all <i>P</i> < .01). Radiomics demonstrated increased intraluminal thrombus (ILT) entropy and reduced ILT density in the endoleak group. A parsimonious model including MMP-9/TIMP-1 ratio and ILT entropy yielded an optimism-corrected area under the curve (AUC) of 0.81 with good calibration, outperforming sac diameter change alone (AUC = 0.62). Persistent endoleak is associated with systemic proteolytic imbalance and imaging-detectable thrombus heterogeneity. Integration of biological and radiomic markers improves risk stratification compared with morphological surveillance alone and warrants prospective validation.Clinical ImpactThis study proposes a novel multimodal approach to post-EVAR surveillance by integrating circulating metalloproteinase inhibitors with AI-derived CT radiomics. Combining biological biomarkers with quantitative analysis of peri-EVAR thrombus volume and density may improve identification of patients at risk of persistent endoleak beyond conventional imaging assessment. For clinicians, this approach could enable earlier risk stratification, support personalised surveillance intervals, and identify patients who may benefit from closer imaging follow-up or intervention. The innovation lies in integrating systemic biological activity with local imaging phenotypes within a predictive model, moving EVAR surveillance from a predominantly anatomical strategy towards a personalised, data-driven biological-imaging paradigm.

Tags

Topics