A total of 11 AI products for intracranial hemorrhage (ICH) management were identified, with 19 supporting scientific publications (Annalise Enterprise CTB, BrainScan CT, CINA-ICH, DeepCT, Deepstroke, Intracranial Hemorrhage (ICH)*, JBS-04K, qER, Rapid ICH, StrokeViewer, Viz ICH). Eleven out of the 19 publications were related to a single AI product, two products had no publications identified. Diagnostic accuracy efficacy (level 2) was addressed by 15 studies (Figure 1). Higher levels (3-6) demonstrating impact on the clinical workflow was addressed by 9 studies.

The AI applications showed diagnostic metrics with the sensitivity ranging from 87.2%  to 97.84%, the specificity ranging from 84.4% to 99.52% and improvement in decision-making processes when tools assisted radiologists. Two studies examined the volumetric quantification (VQ) function, reporting an intraclass correlation coefficient (ICC) of 0.96 [CI: 0.88-0.98] for the qER application and a correlation coefficient r of 0.983 for the Rapid ICH application.

Diagnostic performance demonstrated high sensitivity ranging from 87.2% to 97.84% and specificity from 84.4% to 99.52%. Workflow efficiency showed mixed results, with significant improvements in scan view delay (SVD) and queue-adjusted wait time (QAWT) in some cases (Table 1). Two studies demonstrated reduced patient length of stay (LOS) by up to 1.3 days. Only two articles used European data (Figure 2). Ten articles declared no conflict of interest.