Integration of Hematoma Volume and Density Heterogeneity via the Harmonic Mean Is Associated With Improved Early Prediction of Hematoma Progression in Frontal Lobe Contusion: A Multicenter Retrospective Cohort Study

<title>Abstract</title> <p> <bold>Objective</bold> : Early hematoma progression (HPC) following frontal lobe contusion (FCL) is closely associated with secondary neurological deterioration and unfavorable outcomes. This study aimed to develop a quantitative noncontrast CT‑based composite imaging marker—the volume‑density integration index (VDII) ,constructed using the harmonic mean of baseline contusion volume and density coefficient of variation—and to evaluate its association with early HPC and incremental predictive value, as well as its incremental contribution beyond traditional imaging parameters. <bold>Methods:</bold> This multicenter retrospective cohort study included a development cohort of 341 patients with FCL from the General Hospital of Ningxia Medical University and an external validation cohort of 273 patients with FCL from The Second Affiliated Hospital, Hengyang Medical School, University of South China. Baseline hematoma volume (V) and density coefficient of variation (DCV) were extracted using 3D Slicer. VDII was constructed as the harmonic mean of V and DCV. Four logistic regression models were developed: VDII‑only; a raw component model (Model 1: volume + DCV + clinical laboratory variables); an integrated model (Model 2: VDII replacing volume and DCV while retaining other variables); and an extended model (Model 3: Model 2 + blood glucose + albumin). Model performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration, Brier score, decision curve analysis (DCA), and bootstrap internal validation. <bold>Results:</bold> HPC occurred in 29.0% (training set), 28.3% (test set) and 21.0% (external set). . In the testing set, VDII‑only showed higher discrimination than volume‑only (AUC 0.745 vs. 0.514, P \&lt; 0.001) and DCV‑only (AUC 0.745 vs. 0.659, P = 0.002). In external validation, Model 2 achieved an AUC of 0.763 (95% CI: 0.698–0.824), , compared with 0.699 (95% CI: 0.627–0.769) for Model 1; this difference was statistically significant after Bonferroni correction for multiple pairwise comparisons (P \&lt; 0.001). . Model 3 did not significantly improve discrimination compared with Model 2 (AUC 0.758 vs. 0.763, P = 0.649).. . In multivariable analysis,lower standardized VDII was independently associated with higher risk of HPC (odds ratio [OR] 0.425, 95% confidence interval [CI]:0.269–0.671, P&lt;0.001). Restricted cubic spline analysis suggested a nonlinear association between VDII and HPC risk (P for nonlinearity = 0.0007) Subgroup analysis showed that VDII had stronger predictive value in small‑volume lesions (P for interaction = 0.005). <bold>Conclusion</bold> : VDII, a harmonic mean‑based imaging composite imaging marker integrating hematoma volume and density heterogeneity, was associated with early HPC in patients with FCL and showed better discrimination than volume or DCV alone in the testing set. Replacing raw volume and DCV with VDII improved external discrimination compared with the raw component model in this retrospective cohort, whereas adding blood glucose and albumin did not provide a significant incremental benefit. These findings suggest that VDII may help early risk stratification, particularly in patients with small initial contusions; however, prospective validation and automated measurement workflows are needed before routine clinical implementation.. </p>