Back to all papers

Small Intracerebral Hemorrhage: Baseline CT Edema-to-Hematoma Ratio as an Independent Predictor of Early Expansion.

February 25, 2026pubmed logopapers

Authors

Shi N,Zhang Y,Ma Y,Zhao H

Affiliations (2)

  • From the Department of Radiology (N.S., Y.Z., H.Z.), The Second Affiliated Hospital of Jiaxing University, Jiaxing 314000, China and Zhejiang Chinese Medical University (Y.M.), Hangzhou 310000, China.
  • From the Department of Radiology (N.S., Y.Z., H.Z.), The Second Affiliated Hospital of Jiaxing University, Jiaxing 314000, China and Zhejiang Chinese Medical University (Y.M.), Hangzhou 310000, China. [email protected].

Abstract

Small intracerebral hemorrhage, defined as baseline noncontrast CT (NCCT) hematoma volume <30 mL, is often considered lower risk for early expansion. We tested whether the edema-to-hematoma ratio (EHR) on baseline noncontrast CT (NCCT) independently predicts expansion and whether combining EHR with hematoma volume (HV) enables practical bedside risk stratification. We retrospectively analyzed 219 conservatively managed patients with small-volume intracerebral hemorrhage. Baseline NCCT underwent AI-assisted 3D volumetry to quantify HV and perihematomal edema and to compute EHR. The primary outcome was binary early hematoma expansion on the first follow-up CT. Multivariable logistic regression included HV, EHR, age, deep location, intraventricular hemorrhage, and anticoagulation status. Discrimination and calibration were assessed with five-fold cross-validation. Youden-index cut points were used to derive a simple bedside rule combining HV and EHR. Early expansion occurred in 20.1% (44/219). Lower EHR and larger HV independently predicted expansion. Per 0.10 increase in EHR, adjusted odds decreased by 23% (aOR, 0.77; 95% CI, 0.66-0.88); per additional 5 mL in HV, odds increased by 37% (aOR, 1.37; 95% CI, 1.09-1.73). A bedside rule (HV ≥9.2 mL plus EHR ≤0.683) identified a high-risk phenotype with a 39.3% expansion rate versus 4.4% in the low-risk reference. The multivariable model outperformed either predictor alone (AUC, 0.748 vs 0.700 and 0.672) with acceptable calibration (intercept, -0.17; slope, 0.92; Brier score, 0.141). A lower baseline EHR independently predicts early expansion in small-volume intracerebral hemorrhage. Pairing EHR with HV yields a practical bedside rule with a large risk gradient and improves discrimination beyond volume alone, supporting targeted monitoring and timely repeat imaging.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.