Background: There is increasing interest in forecasting postoperative complications using bone density metrics. Vertebral Hounsfield unit measurements obtained from CT scans performed for surgical planning or other purposes, known as opportunistic CTs, have shown promise for their ease of measurement and the ability to target density measurement to a particular region of interest. Concomitant with the rising interest in prognostic bone density measurement use has been the increasing adoption of intraoperative advanced imaging techniques. Despite the interest in both outcome prognostication and intraoperative advanced imaging, there is little information regarding the use of CT-based intraoperative imaging as a means to measure bone density.
Questions/purposes: (1) Can vertebral Hounsfield units be reliably measured by physician reviewers from CT scans obtained intraoperatively? (2) Do Hounsfield units measured from intraoperative studies correlate with values measured from preoperative CT scans?
Methods: To be eligible for this retrospective study, patients had to have been treated with the use of an intraoperative CT scan for instrumented spinal fusion for either degenerative conditions or traumatic injuries between January 2015 and December 2022. Importantly, patients without a preoperative CT scan of the fused levels within 180 days before surgery or who were indicated for surgery because of infection, metastatic disease, or who were having revision surgery after prior instrumentation were excluded from the query. Of the 285 patients meeting these inclusion criteria, 53% (151) were initially excluded for the following reasons: 36% (102) had intraoperative CT scans obtained after placement of instrumentation, 16% (47) had undergone intraoperative CT scans but the studies were not accessible for Hounsfield unit measurement, and 0.7% (2) had prior kyphoplasty wherein the cement prevented Hounsfield unit measurement. Finally, an additional 19% (53) of patients were excluded because the preoperative CT and intraoperative CT were obtained at different peak voltages, which can influence Hounsfield unit measurement. This yielded a final population of 81 patients from whom 276 preoperative and 276 intraoperative vertebral Hounsfield unit measurements were taken. Hounsfield unit data were abstracted from the same vertebra(e) from both preoperative and intraoperative studies by two physician reviewers (one PGY3 and one PGY5 orthopaedic surgery resident, both pursuing spine surgery fellowships). For a small, representative subset of patients, measurements were taken by both reviewers. The feasibility and reliability of Hounsfield unit measurement were then assessed with interrater reliability of values measured from the same vertebra by the two different reviewers. To compare Hounsfield unit values from intraoperative CT scans with preoperative CT studies, an intraclass correlation using a two-way random effects, absolute agreement testing technique was employed. Because the data were formatted as multiple measurements from the same vertebra at different times, a repeated measures correlation was used to assess the relationship between preoperative and intraoperative Hounsfield unit values. Finally, a linear mixed model with patients handled as a random effect was used to control for different patient and clinical factors (age, BMI, use of bone density modifying agents, American Society of Anesthesiologists [ASA] classification, smoking status, and total Charlson comorbidity index [CCI] score).
Results: We found that Hounsfield units can be reliably measured from intraoperative CT scans by human raters with good concordance. Hounsfield unit measurements of 31 vertebrae from a representative sample of 10 patients, measured by both reviewers, demonstrated a correlation value of 0.82 (95% CI 0.66 to 0.91), indicating good correlation. With regard to the relationship between preoperative and intraoperative measurements of the same vertebra, repeated measures correlation testing demonstrated no correlation between preoperative and intraoperative measurements (r = 0.01 [95% CI -0.13 to 0.15]; p = 0.84). When controlling for patient and clinical factors, we continued to observe no relationship between preoperative and intraoperative Hounsfield unit measurements.
Conclusion: As intraoperative CT and measurement of vertebral Hounsfield units both become increasingly popular, it would be a natural extension for spine surgeons to try to extract Hounsfield unit data from intraoperative CTs. However, we found that although it is feasible to measure Hounsfield data from intraoperative CT scans, the obtained values do not have any predictable relationship with values obtained from preoperative studies, and thus, these values should not be used interchangeably. With this knowledge, future studies should explore the prognostic value of intraoperative Hounsfield unit measurements as a distinct entity from preoperative measurements.
Level of evidence: Level III, diagnostic study.
Copyright © 2024 by the Association of Bone and Joint Surgeons.