To develop an Artificial Intelligence (AI) based anterior circulation Acute Ischemic Stroke (AIS) prognostication model using clinical features, stroke volume and cerebral collaterals.

Prognostication after AIS is usually based on the clinician’s expertise and there is poor utilization of stroke prognostication scales. We aim to harness the power of AI in predicting outcomes after AIS at 3 months using clinical features, stroke volume and cerebral collaterals.

This longitudinal observational analytic study from a University hospital in western India included all patients presenting within 5 days of onset of AIS in anterior circulation from October 2022 to October 2024. Each data set consisted of clinical features; stroke volume (3-D Slicer on MRI DWI images); cerebral collaterals (on CT Angiography using Tan, Maas, Miteff and MCTA scoring systems); Clot Burden Score (CBS annotated using ITK SNAP) and outcomes by modified Rankin Score (mRS) at 3 months.

A total of 166 patients were studied; 109 males (65.7%) with mean age of 58.65 years. Median NIHSS was 6 (IQR 3-10); Median stroke volume was 9.46 cc IQR 2.2-9.1 cc and median CBS was 19 (IQR 17-20). MRS was found to correlate with the NIHSS (p value 0.07) and CBS (Pearson’s correlation 0.2, p value 0.011). MRS was not found to correlate with Stroke volume (Pearson’s correlation 0.018, p value 0.815).  AI models were developed and tested based on Bagging classifier (F1 Score 92%), X G Boost Classifier (F1 Score 93.2%), Decision Tree Classifier (F1 Score 90.5%), Gradient Boosting Classifier (F1 Score 92%), LGBM Classifier (F1 Score 92%), Ada Boost Classifier (F1 Score 90.8), and Random Forest Classifier, which was found to be the most accurate (Accuracy 87.9%, F1 Score of 92.3%, and AUC ROC of 79.3%).