Supplemental Material for: Improving Accuracy of Intracerebral Hemorrhage Volume Measurement Using Three-Dimensional Modeling and Physical Displacement

In this retrospective feasibility study, 21 non-contrast head computed tomography (CT) studies showing intracranial hemorrhages from two publicly available datasets were converted from segmentation masks into stereolithography files and fabricated as 3D printed phantoms. Hematoma volumes were calculated using manual ABC/2 estimations and custom software segmentation volumetry, with Bland-Altman analysis used to assess interrater agreement.

These were compared to 3D phantom volumetry with method comparison using the Wilcoxon signed-rank test and correlation using Spearman coefficients. <br>Results: Median hematoma volume by ABC/2 was 4.4 mL (IQR 1.4-12) mL, segmentation software derived volume was 4.39 mL (IQR 2.1-5.94) mL, and 3D phantom volumetry was 4.0 mL (IQR 2-5) mL. There was no significant difference between ABC/2 measurement and 3D phantom volumetry (p=.16) with weak correlation (Spearman rho=0.39; p=.08) whereas segmentation derived volumes differed from 3D phantom measurements (p=.01) with a strong correlation (Spearman rho = 0.97; p<.001).

Correlation between ABC/2 and segmentation volumetry was also weaker (rho=0.41; p=.06). <br>Conclusions: 3D printed ICH phantoms are a feasible and reproducible platform for validating hemorrhage volumetry. Phantom-determined volumes showed closer alignment with segmentation-based measurements than with ABC/2 measurements, supporting their use as a reference standard for evaluating current and emerging ICH volume measurement techniques.</p>