Deep-learning-derived neuroimaging biomarkers of sarcopenia as predictors of outcome in endovascular thrombectomy in large vessel occlusion acute ischemic stroke

Introduction Sarcopenia is an emerging marker of biological health and is associated with poor outcomes in many disease states. In this study, we aimed to evaluate the utility of muscle biomarkers in predicting clinical outcomes for patients with large vessel occlusion (LVO) acute ischemic stroke (AIS). Methods This was a single-center observational cohort study of consecutive patients that underwent endovascular thrombectomy (EVT) for LVO AIS. A deep-learning model was employed to segment and measure the volume, surface area, and maximum thickness of temporalis and sternocleidomastoid (SCM) muscles. The primary outcome was poor functional outcome, defined by an mRS of 3-6 at 3 months post-stroke. Univariable and multivariable logistic regression models were performed to evaluate associations between the muscle biomarkers and outcome measures after adjusting for clinical variables of age, sex, and NIHSS. Results A total of 297 patients were included. 175 (58.9%) had poor functional outcomes at 3 months post-stroke. For each 10cm 3 decrease in SCM volume (SV) and temporalis volume (TV), the odds of poor functional outcome at 3 months post-stroke increased by 34% (OR 0.66, 95% CI 0.52-0.84, p < 0.001) and 18% (OR 0.82, 95% CI 0.73-0.91, p < 0.001), respectively. After adjusting for age, sex and NIHSS, our baseline outcome model yielded an AUC of 0.716. Including sarcopenia biomarkers in the model improved discrimination: SV dichotomized (adjusted OR (aOR) 0.39, 95% CI 0.20-0.74, p-value <0.01, AUC: 0.731), TV dichotomized (aOR 0.51, 95% CI 0.30-0.86, p-value 0.012, AUC: 0.724). Conclusion Our study identified that temporalis and SCM muscle volumes were independently associated with functional outcomes after EVT for LVO AIS.