Development and validation of a machine learning model for predicting intracranial atherosclerotic disease in large vessel occlusion prior to endovascular therapy.

BACKGROUND AND PURPOSE: Early identification of large vessel occlusion due to underlying intracranial atherosclerotic disease (ICAD-LVO) in acute ischemic stroke (AIS) is crucial to determine optimal treatment strategies. We aimed to develop and evaluate machine learning (ML) models to predict ICAD-LVO based on variables, including basic patient information, medical history, neurological findings, and imaging data, which are commonly assessed in stroke management prior to endovascular therapy (EVT).

METHODS: This study is a retrospective cohort study using data from the prospective multicenter registry, RESCUE-Japan Registry 2, conducted between October 2014 and September 2016. Patients with intracranial internal carotid artery or middle cerebral artery occlusion who underwent EVT were included; those with tandem occlusion, unknown etiology, and missing data were excluded. The data were divided by enrollment period into training and test datasets. ML models including Random Forest (RF) and Extreme Gradient Boosting (XGBoost) were trained and evaluated using metrics such as accuracy, balanced accuracy, area under the curve (AUC), recall, precision, and F1 score.

RESULTS: Of 2420 registered patients, 991 met the inclusion criteria; after division by enrollment period and exclusions, 623 (training) and 191 (test) patients were analyzed. The RF model achieved an accuracy of 0.75, a balanced accuracy of 0.77, and an AUC of 0.87. The XGBoost model achieved an accuracy of 0.76, a balanced accuracy of 0.80, and an AUC of 0.87.

CONCLUSIONS: The ML models effectively predicted ICAD-LVO using variables commonly assessed before EVT, enabling pre-procedural inference of LVO etiology and facilitating individualized treatment strategies.