acemachine learning prediction of hematoma expansion, poor functional outcome, and mortality in spontaneous intracerebral hemorrhage
Source article: Machine Learning-Based Prediction of Poor Outcomes in Intracerebral Hemorrhage: A Systematic Review and Meta-Analysis
Spontaneous intracerebral hemorrhage (ICH) is associated with high risks of mortality and disability, yet early and accurate outcome prediction remains challenging. This study systematically evaluated the performance of machine learning (ML) models in predicting key adverse outcomes (hematoma expansion [HE], poor functional outcome, mortality) in ICH, aiming to provide consolidated evidence for future research and clinical translation. We systematically searched PubMed, Embase, Web of Science, and Cochrane Libra…
Positive state: both sides are scored from claims and sources, not community votes.
Intracerebral by Lucien Monfils. CC BY-SA 3.0 · https://creativecommons.org/licenses/by-sa/3.0
A systematic review and meta-analysis up to September 2025 synthesized 83 studies involving at least 136,840 patients to evaluate machine learning models predicting hematoma expansion, poor functional outcome, and mortality after spontaneous intracerebral hemorrhage. Pooled analyses found that models combining clinical and radiomics features achieved the highest discrimination, with C-indexes of 0.822, 0.850, and 0.860 respectively, largely from internal validation sets.
High discrimination suggests potential to improve early risk stratification for a condition with high mortality and disability, but the evidence base as of the 2026 publication date is constrained by reliance on internal validation and sparse external validation. The comparable performance of logistic regression to more complex algorithms also raises questions about added value, and clinical translation will require validation in diverse, prospective cohorts.
- Systematic review and meta-analysis included 83 studies involving at least 136,840 patients with spontaneous intracerebral hemorrhage, searching PubMed, Embase, Web of Science, and Cochrane Library up to September 2025.
- Pooled C-index from predominantly internal validation was 0.822 for hematoma expansion, 0.850 for poor functional outcome defined as modified Rankin Scale 3-6, and 0.860 for mortality for clinical-radiomics models.
- Logistic regression exhibited performance comparable to more complex ML algorithms across prognostic tasks.
- Authors noted findings should be interpreted with caution due to sparse true external validation.
Machine learning models that integrate clinical and radiomics features achieved high pooled discrimination for predicting hematoma expansion, poor functional outcome, and mortality in adults with spontaneous intracerebral hemorrhage.
Pooled performance estimates were based predominantly on internal validation, with true external validation remaining sparse, limiting confidence in generalizability.
The rundown
The review included 83 studies with at least 136,840 adults with spontaneous ICH, evaluating ML models for three adverse outcomes: hematoma expansion, poor functional outcome defined as modified Rankin Scale 3-6, and mortality. Pooled C-index, sensitivity and specificity were calculated using random-effects or bivariate models, with most data coming from internal validation.
Integrated clinical-radiomics models had the highest discrimination, with pooled C-index 0.822 for HE, 0.850 for poor functional outcome, and 0.860 for mortality. Logistic regression performed comparably to more complex algorithms. The authors concluded models hold potential to enhance risk stratification and guide personalized management pending further validation in diverse cohorts.
Performance estimates were derived predominantly from internal validation sets and true external validation remains sparse, limiting generalizability to diverse cohorts.
Sources
- Peer-reviewedBrain and Behavior2026-08-01
How should this claim be treated?
The debate