E2342. Automated CT-Based Imaging Sarcopenia Biomarker Associated with Functional Independence in Acute Ischemic Stroke Patients
  1. Xi Zhen Low; National University Hospital, Singapore
  2. Andrew Makmur; National University Hospital, Singapore
  3. Kevin Teo; Yong Loo Lin School of Medicine, National University of Singapore
  4. Yao Neng Teo; Yong Loo Lin School of Medicine, National University of Singapore
  5. Yilei Wu; Yong Loo Lin School of Medicine, National University of Singapore
  6. Yichi Zhang; Yong Loo Lin School of Medicine, National University of Singapore
  7. Leonard Yeo; National University Hospital, Singapore
More than half of acute ischemic stroke (AIS) patients with large vessel occlusions have significant disability despite successful recanalization from endovascular therapy (EVT). Sarcopenia is an emerging marker of biological health and is strongly associated with poor outcomes in many disease states, including ischemic stroke. We hypothesise that quantitative measurement of the temporalis muscle, a surrogate biomarker for sarcopenia, developed using a deep learning model trained on routine CTA of the brain will predict clinical outcomes after AIS.

Materials and Methods:
Consecutive patients who underwent EVT for large vessel occlusion AIS from a single comprehensive stroke centre between 2016 and 2019 were included. Patients underwent multidetector CTA from the aortic arch to the cranial vertex with 1 mm cuts. All scans were extracted and annotated by a team of neuroradiologists to train a deep learning model to segment the temporalis muscle to produce the following parameters - temporalis volume (TV), temporalis surface area (TSA), and temporalis maximum thickness (TMT). Traditional clinical variables including age, comorbidities and stroke severity measured using NIHSS were collected. The primary outcome was defined as functional independence (FI), defined by a modified Rankin scale (mRS) of 0 - 2 assessed at 90 days. Univariate analyses were performed using Student’s t-test to determine the association of the TV, TSA and TMT against the primary outcome. AIS patients were stratified into two groups based on their temporalis muscle parameters (<25th percentile vs. equal to the 25th percentile) for further analyses. Logistic regression was performed to evaluate independent associations with the primary outcome, after adjusting for traditional clinical variables.

A total of 322 AIS patients were included with a median age of 67.5 years (IQR 57-75); 146 patients (45%) were female. 121 patients (37.5%) achieved FI (mRS 0-2) at 90 days. Patients who achieved FI had higher TV (7584.55 mm3 vs.6764.14 mm3, p=<0.02). Differences in TSA (2174.53 mm2 vs. 2067.77 mm2, p=0.07) and TMT (14.55 mm vs 14.02 mm, p= 0.08) were not statistically significant. AIS patients with TV<25th percentile achieved a significantly lower rate of FI compared to patients with TV=25th percentile (23.5% vs. 42.3%, OR 0.42, 95% CI 0.24-0.74, p=0.002). On multivariable analyses, TV=25th percentile remained independently associated with FI (OR 3.59, 95%CI 1.43 - 9.00,p=0.006). A combined model including TV=25th percentile, age and NIHSS yielded an AUC of 0.725 (0.667 to 0.783) compared to NIHSS (AUC = 0.679, 0.617 to 0.740) and age (AUC = 0.660, 0.599 to 0.721) alone.

In this AIS cohort, temporalis muscle volume set at a cut-off of =25th percentile obtained on routine CTA imaging was independently associated with good functional outcome after EVT. Deep learning and automation allows for reliable and rapid assessment of these biomarkers, which may aid clinicians in patient selection for EVT. Further prospective studies are necessary to validate our findings.