Short Oral + Poster Presentation Asia Pacific Stroke Conference 2024

Prediction of thrombectomy outcomes using perfusion profile of eloquent brain regions: A multicenter study (107171)

Haipeng Li 1 , Ho Ko 1 , Sangqi Pan 1 , Lt Lui 2 , Trista Hung 1 , Chun Ngo Yau 1 , Edward Hui 3 , Xinyi Leng 2 , Jill M. Abrigo 3 , Bonnie YK Lam 1 , Vincent Ct Mok 1 , Rosa HM Chan 2 , Wai Hong Thomas Leung 1 , Hao Wang 4 , Yuan Che Feng 4 , Bonaventure YM Ip 1
  1. Division of Neurology, Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Hong Kong, China
  2. Department of Electrical Engineering, Faculty of Engineering, City University of Hong Kong, Hong Kong, China
  3. Department of Imaging and Interventional Radiology, Faculty of Medicine, The Chinese University of Hong Kong, Hong Kong SAR, China
  4. Department of Neurology, Linyi People’s Hospital, Linyi, Shandong Province, China

Background/Aims

Pivotal endovascular thrombectomy (EVT) for acute large vessel occlusion (LVO) based on ischemic core and core-penumbra mismatch were suboptimal in predicting EVT outcomes. Hypothesizing that the perfusion status of eloquent brain regions is more crucial in post-EVT prognostication, we aimed to determine EVT outcomes in patients with critical hypoperfusion in pre-specified eloquent brain regions.

Methods

In this multicentre retrospective study, we retrieved patients with acute middle cerebral artery or internal carotid artery occlusion who received CT perfusion before EVT from 4 hospitals in Hong Kong and mainland China in recent 4 years, excluding patients with angiographic results of mTICI<2b. We automatically generated perfusion profiles for 79 brain regions using a custom-developed pipeline, and assessed the association between the predictor of regional ischemic core to regional hypoperfused area ratio (unit in 10%) and poor functional recovery (mRS>3) using multivariate logistic regression adjusted for ischemic core, ASPECTS, collateral score, and other clinical indicators. 

Results

Among 310 identified acute LVO patients, 83 (26.8%) had poor functional recovery despite successful EVT. Multivariate logistic regression revealed the predictor in the primary motor cortex (aOR=1.38, p=0.015), lentiform nucleus (aOR=1.16, p=0.038), internal capsule (aOR=1.23, p=0.029), caudate (aOR=1.17, p=0.018) and the eloquent involvement of these regions (aOR=1.72, p<0.01) were associated with poor functional recovery independent of other control factors. The mean absolute SHapley Additive exPlanations on the eloquent regions show a strong predictive power (SHAP=0.23) to the poor clinical outcome. 

Conclusion

The predictive model using the eloquent regions significantly improves the prognostic accuracy of successful EVT.

  1. Billot, B., Magdamo, C., Cheng, Y., Arnold, S. E., Das, S., & Iglesias, J. E. (2023). Robust machine learning segmentation for large-scale analysis of heterogeneous clinical brain MRI datasets. Proceedings of the National Academy of Sciences - PNAS, 120(9), e2216399120–e2216399120. https://doi.org/10.1073/pnas.2216399120
  2. Ruogu Fang, Shaoting Zhang, Tsuhan Chen, & Sanelli, P. C. (2015). Robust Low-Dose CT Perfusion Deconvolution via Tensor Total-Variation Regularization. IEEE Transactions on Medical Imaging, 34(7), 1533–1548. https://doi.org/10.1109/TMI.2015.2405015