RESUMO
STUDY DESIGN: Retrospective case series. OBJECTIVE: To identify specific magnetic resonance imaging (MRI) characteristics of epidural fluid collections associated with infection, hematoma, or cerebrospinal fluid (CSF). SUMMARY OF BACKGROUND DATA: Interpretation of postoperative MRI can be challenging after lumbar fusion. The purpose of this study was to identify specific MRI characteristics of epidural fluid collections associated with infection, hematoma, or CSF. METHODS: The study population includes consecutive patients between 2006 and 2010 who had MRIs performed within 2 weeks after elective surgery for evaluation of possible CSF fluid collection, hematoma, or infection. Patients with known previous infection (discitis/osteomyelitis) or inadequate MRIs were excluded from the study. Medical records were reviewed to determine the diagnosis (infection, hematoma, or pseudomeningocele) underlying the fluid collection. MRIs were retrospectively evaluated by a musculoskeletal radiologist and orthopedic spine attending who were blinded to the pathologic diagnosis for characteristics of the fluid collection. MRI characteristics include location of lesion: osseous involvement, disk location, anterior versus posterior versus anteroposterior, soft-tissue involvement, and iliopsoas involvement. Characteristics of the lesion include: volume of lesion, loculation, satellite lesions, multiple loci, destructive characteristics, and mass effect upon thecal sac. Enhancement was scored based upon the following variables: rim enhancement, smooth versus irregular, thin versus thick, heterogeneity, diffuse enhancement, nonenhancement, and rim thickness. General fluid collection intensity and complexity on T1, T2, and T1 postcontrast images was scored as high, medium, and low. The χ test was used to compare the incidence of imaging characteristics between patient groups (infection, hematoma, and CSF). RESULTS: Thirty-three patients were identified who met inclusion criteria. There were 13 (39%) with infection, 9 (27%) with hematoma, and 11 (33%) with CSF collection. Factors that were associated with infection were osseous involvement (R=0.392, P=0.024) and destructive characteristics (R=0.461, P=0.007). Factors that were correlated with hematoma include mass effect (R=0.515, P=0.002) and high T1-signal intensity (R=0.411, P=0.019), absence of thecal sac communication (R=-0.389, P=0.025), and absence of disk involvement (R=-0.346, P=0.048). Pseudomeningocele was associated with thecal sac communication (R=0.404, P=0.02), absence of mass effect (R=-0.48, P=0.005), low T1 signal (R=-0.364, P=0.04), and low T2 complexity (R=-0.479, P=0.005). CONCLUSION: Specific characteristics of the postoperative MRI can be used to distinguish infection from noninfectious fluid collections. The strongest predictors of infection were osseous involvement and destructive bony changes. Hematoma was associated with mass effect on the thecal sac, high T1-signal intensity, and absence of thecal sac communication and disk involvement. CSF collections were distinguished by absence of mass effect, low T2-signal complexity, low T1-signal intensity, and communication with the thecal sac.