Table 3.

The clinical and imaging variables used to train the prediction models.

Clinical and demographic information	Imaging-based clinical features	Vascular geometric features
Age	ASPECTS score	Total length
Sex	Side of the stroke	Number of branches
Baseline NIHSS	Occlusion location	Total volume
History of stroke	Perfusion mismatch ratio and volume	Average diameter
Smoker (y/n)	Collateral index (CI)—ground truth	Tortuosity
Diabetes mellitus (y/n)		Fractal dimension
Hypertension (y/n)		Auto-estimated CI

NIHSS, National Institute of Health Stroke Scale; ASPECTS, Alberta Stroke Program Early CT Score. The cells in blue represent the conventionally used features, yellow refers to the automatically estimated CI using our method and green cells represent the novel vascular predictors.