Size and location of intracranial hemorrhage in acute stroke presentations: correlation with baseline NIHSS and relevance to direct-to-angiography workflows.

BACKGROUND AND PURPOSE: Direct-to-angiography (DTA) workflows for suspected large vessel occlusion stroke reduce treatment delays by bypassing emergency-department imaging, but their safety depends on excluding intracranial hemorrhage (ICH) on flat-panel detector CT (FDCT), where posterior fossa visualization is limited by beam-hardening artifacts. Epidemiological data on ICH size and location relative to stroke severity are limited. We characterized ICH type, volume, and location in acute stroke presentations and their relationship with baseline NIHSS. MATERIALS AND METHODS: We retrospectively analyzed 394 patients with imaging-confirmed spontaneous ICH from the prospective Swiss Stroke Registry at a high-volume comprehensive stroke center. Hemorrhages were classified by type, location, and volume using semi-automated segmentation. Associations with baseline NIHSS were assessed using Spearman's correlation, Wilcoxon rank-sum tests (NIHSS <7 vs. ≥7), and generalized additive and ordinal regression models adjusted for age, location, and secondary extension. RESULTS: Intraparenchymal hemorrhage accounted for 89% of cases, followed by subarachnoid (7%), intraventricular (3%), and subdural (<1%). Median volume was 18 mL (IQR 5-51). Volume correlated with NIHSS (Spearman ρ = 0.56, 95% CI 0.48-0.64, p < 0.001), with a significant nonlinear association in adjusted models. Patients with NIHSS ≥7 had larger volumes than those with NIHSS <7 (median 38 vs. 6 mL, p < 0.001). Lobar hemorrhages had the largest volumes. In contrast, infratentorial hemorrhages, particularly brainstem bleeds, produced severe deficits at small volumes (median 6 mL) and were not associated with larger volumes after adjustment, indicating that location influences severity independently of size. Isolated infratentorial hemorrhage accounted for only 2.8% of NIHSS ≥7 patients. CONCLUSION: Severity in supratentorial ICH is volume-dependent, whereas location, particularly brainstem involvement, drives severity in isolated infratentorial ICH independently of volume. Although rare among high-NIHSS patients, these small-volume bleeds may be hardest to detect on FDCT and warrant attention to posterior fossa image quality in DTA workflows. Because detectability was not measured, prospective paired FDCT-MDCT studies are needed.