🧠 A new brain-hemorrhage prognostic score is a useful reminder: in stroke, when you measure can matter as much as what you measure.
Predicting outcomes after an intracerebral hemorrhage (ICH) is still one of the harder calls in acute neurology. Families want answers early, and clinicians are often working from a single CT snapshot plus a few clinical variables. The established risk scores help, but most land around an AUC of 0.72–0.81.
A study published this week in Annals of Clinical and Translational Neurology (June 25, 2026) introduces HAGIV, a point-based score built to narrow that gap. The interesting part isn't the acronym. It's one of the inputs.
Alongside the familiar predictors (hematoma volume, age, Glasgow Coma Scale score, and intraventricular hemorrhage), HAGIV adds a time-adjusted measure of imaging instability: how the bleed is behaving over time on non-contrast CT, not just how large it looks on the first read.
The reported results: 📊 AUC 0.86 in derivation and 0.84 in validation, outperforming four established ICH scores, with the differences statistically significant.
This was a retrospective, multicenter cohort, and the authors are refreshingly upfront about the limits: internal validation only, a mostly small-to-moderate supratentorial ICH population, and an explicit call for prospective external validation before anyone relies on it at the bedside. The abstract also doesn't report the cohort size or confidence intervals, so those point estimates deserve context rather than a victory lap.
That mix of a genuinely interesting idea and honest reporting of its limits is what strong prognostic-model research should look like.
The broader signal for our field: prognosis is shifting from static snapshots toward models that read the trajectory of an injury. That's harder to validate, and far more useful when it holds up.
One question I keep coming back to: how much should an early, not-yet-externally-validated score shape prognostic conversations with families? Curious how neuro-ICU and stroke teams are drawing that line.
#StrokeAI #Neuroradiology #BrainHemorrhageDetection #Neurology #ImagingIT