The prediction of hematoma growth in acute intracerebral hemorrhage: from 2-dimensional shape to 3-dimensional morphology

INTRODUCTION: The relationship between the 3-dimensional morphological features of hematoma and hematoma growth (HG) remains unclear. We aimed to quantitatively assess the predictive value of 3-dimensional hematoma morphology for HG among patients with intracerebral hemorrhage (ICH). METHODS: Our study comprised 312 consecutive ICH patients. Using semi-automated volumetric analysis software, we measured hematoma volumes and delineated the region of interest. We employed Python software to extract shape features and receiver operating characteristic curve analysis to assess the predictive performance of hematoma morphology for HG. p < 0.05 was considered statistically significant. RESULTS: Sphericity and SurfaceArea emerged as the most effective 3-dimensional hematoma morphological predictors for HG. Optimal cutoff points relating to HG were Sphericity ≤0.56 and SurfaceArea >55 cm2. We subsequently constructed the 3-dimensional morphology models, including the probability of hematoma morphology (PHM) and the probability of comprehensive model (PCM), to predict HG. The PHM model outperformed the irregular hematoma (p = 0.007), island sign (p = 0.032), and satellite sign (p < 0.001) in predictive accuracy for HG. Among all prediction models, the PCM presented the highest predictive value for active bleeding. CONCLUSIONS: The Sphericity ≤0.56 and SurfaceArea >55 cm2 could represent the optimal threshold for HG prediction. PHM was considered a reliable 3-dimensional morphology model for HG prediction. PCM tended to be a better model for risk stratification of active bleeding in acute ICH patients.