P-029 Interpretable machine learning modeling for thrombectomy outcome prediction in ischemic stroke. (23rd July 2022)

• Record Type: Journal Article
• Authors: Jabal, M; Joly, O; Kallmes, D; Harston, G; Rabinstein, A; Huynh, T; Brinjikji, W
• Is Part Of: Journal of neurointerventional surgery. Volume 14(2022)Supplement 1
• Page Start: A67
• Page End: A67
• Publication Date: 2022-07-23
• Subjects: Nervous system -- Surgery -- Periodicals; Cerebrovascular disease -- Surgery -- Periodicals; 617.48
• DOI: 10.1136/neurintsurg-2022-SNIS.101
• Languages: English
• ISSNs: 1759-8478

Abstract

Abstract : Background and Purpose: Mechanical thrombectomy greatly improves stroke outcomes. Yet, some patients fall short of full recovery despite good reperfusion. The goal of the study was to explore the value of machine learning pipelines in predicting functional outcome of acute ischemic stroke in the acute setting upon initial screening at the emergency department prior to intervention, and to identify the patients who would highly benefit if a mechanical thrombectomy procedure were to be performed, using clinical and auto-extractable radiological information consistently available upon first emergency evaluation. With the purpose of stratifying patients with large vessel occlusion stroke prior to the endovascular procedure leading to enhanced acute management decision-making. Materials and Methods: A two-center retrospective cohort of 293 patients with acute ischemic stroke who underwent thrombectomy was analyzed. Machine learning models were developed to predict dichotomized modified Rankin score at 90 days using clinical and imaging features, both separately and combined. Conventional and experimental imaging biomarkers were quantified using automated image processing software from non-contract CT and CTA. Shapley additive explanation was applied for model interpretability and predictor importance analysis of the optimal model. Results: Merging clinical and imaging features returned the best results for mRS-90 prediction. The best performing classifier was Extreme Gradient Boosting with an AUC = 84% using selected features. The most important classifying features were age, baseline NIHSS, occlusion site, degree of brain atrophy (primarily represented by cortical CSF volume and lateral ventricle volume), early ischemic core (primarily represented by e-ASPECTS), and collateral circulation deficit volume on CTA. Conclusion: Machine learning applied to quantifiable image features from CT and CTA alongside basic clinical characteristics constitutes a promising automated method for predicting stroke prognosis. Our interpretable model allows understanding of which features contribute the most to post-thrombectomy outcome prediction and which feature values the model used to individually predict each patient outcome. Disclosures: M. Jabal: None. O. Joly: 5; C; Brainomix. D. Kallmes: None. G. Harston: 4; C; Brainomix. 5; C; Brainomix. A. Rabinstein: None. T. Huynh: None. W. Brinjikji: None.