Computed Tomography Volumetry of Bone Cement: Retrospective Blinded Validation of Commercially Available Semi-automated Edge Detection Software

Alan A Sag; Ernest Barral; Russell E Thompson; C Rory Goodwin; James Ronald

doi:10.1097/RCT.0000000000001588

Computed Tomography Volumetry of Bone Cement: Retrospective Blinded Validation of Commercially Available Semi-automated Edge Detection Software

J Comput Assist Tomogr. 2024 Feb 5. doi: 10.1097/RCT.0000000000001588. Online ahead of print.

Authors

Alan A Sag, Ernest Barral¹, Russell E Thompson², C Rory Goodwin, James Ronald²

Affiliations

¹ Duke University School of Medicine.
² From the Division of Vascular and Interventional Radiology, Department of Radiology.

PMID: 38335943
DOI: 10.1097/RCT.0000000000001588

Abstract

Objective: Cement volumes are increasingly linked to orthopedic oncology and neurosurgical outcomes (construct durability, adjacent fracture), but manual cement volumetry remains time prohibitive. The authors aim to report performance of PACS-integrated volumetric software specifically for barium-enhanced polymethylmethacrylate cement.

Methods: Institutional review board-approved single-institution retrospective review of patients from 2019-2022 undergoing kyphoplasty for pathological compression fractures with a quantitative cement infuser providing true cement volume. An operator blinded to true cement volumes retrospectively performed software-assisted volumetry on follow-up computed tomography scans.

Results: Included were 91 kyphoplasty levels in 56 patients: mean age, 62 years (range, 34-85 years), 73% female. True cement volume (available for 44 of 66 procedures) was mean 4.5 mL per level (range, 1.2-15.6 mL). Measured cement volume (available for all procedures) yielded a mean of 6.1 mL per level (range, 1.5-27.9 mL). For the 57 levels (39 patients) where both true and measured cement volumes were available, linear regression intercept and slope were 1.46 (95% CI = 0.97-1.95, P < 0.001) and 0.52 (CI = 0.47-0.57, P < 0.001), respectively, suggesting measured volume averaged 1.46 mL greater than true volume, with each additional milliliter of measured volume corresponding to approximately 0.52 mL of true volume. There was no significant difference in the relationship between estimated and actual cement volume in thoracic levels (intercept = -0.24, CI = -1.13 to 0.66, P = 0.61; slope = 0.03, CI = -0.14 to 0.19, P = 0.73) compared with lumbar levels. The goodness-of-fit of the regression model was strong (R2 = 0.81). Discrepancies ranged from 90% underestimation to 52% overestimation; average, 17% overestimation.

Conclusions: Semi-automated volumetry maintained a strong correlation with true volumes across the thoracic and lumbar curvatures, overestimating cement volume by a mean of 17% or 1.46 mL.