Published online Jun 16, 2023.
https://doi.org/10.13104/imri.2022.1105
Diagnosis of Infectious Spondylitis Using Non-Contrast Enhanced MRI With Axial Diffusion-Weighted Images: Comparison With Gadolinium-Enhanced MRI
Abstract
Purpose
To assess the diagnostic performances of diffusion-weighted imaging (DWI)-included non-contrast magnetic resonance imaging (MRI) compared to standard contrast-enhanced MRI for infectious spondylitis.
Materials and Methods
This study involved 154 participants: a spondylitis group (n = 76) and a control group (n = 78) with Modic type 1 degeneration or recent compression fractures. Two readers independently reviewed paraspinal soft tissue signal change and abscess with 5-scale confidence scores based on two image sets: one featuring both non-contrast-enhanced MRI (NCEI) and DWI and the other consisting of NCEI and contrast-enhanced fat-suppressed T1-weighted imaging (CEFST1). The diagnostic performance of the two image sets was compared using McNemar tests for sensitivity, specificity, and area under the receiver operating characteristics (AUROC) analysis. Interobserver agreements (κ) for each images sets were also calculated.
Results
The sensitivity and specificity for infectious spondylitis were 90.8% and 69.2% for NCEI + DWI, 96.1% and 60.3% for NCEI + CEFST1 in reader 1, whereas it was 92.1% and 66.7% for NCEI + DWI, and 96.1% and 68.0% for NCEI + CEFST1 in reader 2. Sensitivities and specificities were not significantly different between NCEI + DWI and NCEI + CEFST1 (reader 1: p = 0.289, 0.065; reader 2: p = 0.250, > 0.999, respectively). However, the AUROC was not considerably different between the two modalities in only one reader (p = 0.306 in reader 1, p = 0.031 in reader 2). Interobserver agreement for infectious spondylitis was moderate (κ = 0.55) in NCEI + DWI and substantial (κ = 0.66) in NCEI + CEFST1.
Conclusion
Non-contrast enhanced MRI with additional DWI is as effective for diagnosing infectious spondylitis as a contrast-enhanced MRI.
INTRODUCTION
Infectious spondylitis is an infection of one or more components of the spine, such as the vertebra, intervertebral discs, paraspinal soft tissues, and epidural space [1]. As the infection progresses, the intervertebral discs and vertebral bodies are destroyed, leading to spinal deformity, instability, and decreased neurologic functioning. Therefore, early detection and timely treatment are pivotal to preventing severe complications [2]. Magnetic resonance imaging (MRI) is the modality of choice in diagnosing spondylitis because of its clear demarcation of anatomic structures in the spines and sensitive visualization of inflammatory changes [2, 3]. In particular, gadolinium (GD) enhancement is often required to assess paraspinal soft tissue inflammation and abscess formation, which can indicate infectious spondylitis over other entities such as compression fractures, degenerative bone marrow change, or spondyloarthropathy. However, the use of intravenous GD enhancement carries the risk of nephrogenic systemic fibrosis and immediate hypersensitive reaction [4, 5, 6]. Consequently, physicians face a diagnostic dilemma when patients with renal insufficiency require MRI for spondylitis.
Diffusion-weighted imaging (DWI) offers insight into the diffusion properties of water molecules in tissue [7]. It has been established as a useful tool in detecting abscesses [8, 9] and providing information on superficial and deep soft tissue infection [3, 9, 10], degenerative disease [9, 11], and inflammatory rheumatic disease [9]. However, the diagnostic accuracy of DWI, when combined with conventional magnetic resonance (MR) sequences for infectious spondylitis, has yet to be explored. Therefore, the purpose of our study was to compare the diagnostic performance between DWI combined with non-contrast enhanced MRI and contrast-enhanced MRI for infectious spondylitis.
MATERIALS AND METHODS
This retrospective study was approved by the institutional review board of Seoul St. Mary’s Hospital, College of Medicine, The Catholic University of Korea (IRB No. KC21RISI0652), and complied with Health Insurance Portability and Accountability Act (HIPAA) guidelines. Informed consent was not required.
Patients
The primary aim of this research was to demonstrate the non-inferiority of DWI in comparison to conventional contrast-enhanced MRI for the diagnosis of infectious spondylitis. The sample size was estimated to be 59 in each group based on the expected sensitivity of conventional contrast-enhanced MRI of 90%, a 10% difference in the sensitivity being the non-inferiority margin, and a correlation between paired observations of 0.8. The significance level and statistical power were both set at 0.05. Between January 2018 and April 2021, 96 patients with infectious spondylitis who underwent contrast-enhanced MRI were included. Upon review of the cases, 20 patients had to be excluded due to incomplete image sets (n = 18), a metallic artifact (n = 1), or the patient being a pediatric case (n = 1). Of the 76 patients in the spondylitis group, the causative organism was identified in 33 of them. The tissue sample was collected by surgery or image-guided biopsy in 48 patients for culture and polymerase chain reaction tests. The specific organism identified were Mycobacterium tuberculosis (n = 8), Mycobacterium intracellulare (n = 1), Candida species (n = 2), Escherichia coli (n = 3), Klebsiella pneumoniae (n = 1), Corynebacterium striatum (n = 1), Streptococcus species (n = 3), and Staphylococcus aureus (n = 1). In 28 patients, blood culture revealed the causal organisms: Staphylococcus aureus (n = 3), Enterococcus spp. (n = 4), Escherichia coli (n = 1), Streptococcus species (n = 4), and Granulicatella adiacens (n = 1). For the remaining 43 patients for who the causative organism was not identified, the diagnosis was made based on a combination of the MRI features (e.g., paraspinal soft tissue enhancement and/or abscess formation), laboratory results (e.g., increased erythrocyte sedimentation rate and C-reactive protein, and elevated white blood cells counts), and clinical findings (fever, severe back pain, and neurologic deficit).
As a control group, 94 patients who underwent contrast-enhanced MRI and had marrow signal alteration, which was not associated with spondylitis, were included. We searched the radiology reports from June 2018 to April 2021 for the term ‘Modic type 1 change’ or ‘recent compression fracture’ to find the patients with marrow signal alteration. The search yielded 31 and 86 patients, respectively. One musculoskeletal radiologist (S.Y.L) with a 10-year of experience reviewed the images and medical records for eligibility. The patients were deemed eligible if there had definite marrow signal change in fat-suppressed T2-weighted images and no evidence that the signal alteration was associated with the infectious spondylitis based on medical records and laboratory results. However, 16 patients were excluded due to the following reasons: 1) no apparent marrow signal alteration on MRI (n = 6), 2) incomplete image sets (n = 9), 3) or combined pathologic fracture (n = 1). Finally, 78 patients were included in the control group.
MRI
MRI of the spines was conducted with Magnetom Verio 3.0 T MR systems (Siemens Medical Solution). The imaging protocol included sagittal T1-weighted turbo spin-echo sequences, sagittal and axial T2-weighted turbo spin-echo sequences for non-contrast-enhanced MRI (NCEI); axial DWI by single-shot spin-echo echoplanar acquisition, sagittal fat-suppressed and axial T1-weighted turbo spin-echo sequence for contrast-enhanced fat-suppressed T1-weighted imaging (CEFST1). The axial DWI is routinely included in the contrast-enhanced spine MRI because it was generally ordered to assess conditions in which DWI might be beneficial, such as tumors, infection, or inflammatory diseases. The axial scan range was tailored to the size of the lesson. The detail of the parameters for each sequence are summarized in Table 1. For contrast enhancement, 1 mL per kg of body weight of meglumine gadoterate (Dotarem, Guerbet) or gadobutrol (Gadovist, Bayer Schering Pharma) was injected intravenously.
MRI Analysis
Two independent readers (J.Y.J with 12 years of experience in musculoskeletal radiology and J.M.H, a board-certified radiologist under the musculoskeletal fellowship training) were employed to review all the images. The readers were kept blind to the clinical information.
MR images from the spondylitis and control groups were randomly mixed and reshuffled. Two image sets- NCEI + DWI and NCEI + CEFST1- were created. Each image set was reviewed in a separate session with a one-month washout period placed between the two reviews to prevent recall bias. In the first session, NCEI + DWI was independently reviewed by the two readers. In the second session, NCEI + CEFST1 was reviewed after a random reshuffling of the order of images. MR images were analyzed for the presence of infectious spondylitis and the formation of an abscess. Infectious spondylitis was defined when there was an increased signal intensity or enhancement in the bone marrow and an obvious paraspinal soft tissue signal change or enhancement in axial high b-value DWI or CEFST1, respectively. Abscess formation was defined when there was a well-circumscribed cystic lesion in epidural or paraspinal soft tissue, which demonstrated enhancing wall on axial CEFST1 and high signal intensity on both high b-value DWI and ADC map. Five-level confidence scores ranging from 0; definite negative, 1; probably negative, 2; equivocal, 3; probably positive, 4; definitely positive (Fig. 1) were used for analysis. The confidence scores were dichotomized into 0–2: negative and 3–4: positive to calculate sensitivity, specificity, and interobserver agreements. For the standard reference of abscess, consensus reading between two readers was performed based on NCEI, DWI, and CEFST1 after completing the independent image analysis.
Fig. 1
Representative images of each confidence level in the magnetic resonance analysis. Five-level confidence score system ranges from 0 to 4 according to the readers' confidence level on the presence of spondylitis. 0; definite negative, 1; probably negative, 2; equivocal, 3; probably positive, 4; definitely positive were used for analysis.
Statistical Analysis
Statistical significance was established with a two-sided p-value of less than 0.05 for the aforementioned statistical tests. The sensitivity and specificity for diagnosing infectious spondylitis and abscess were calculated for NCEI + DWI and NCEI + CEFST1, and McNemar tests were performed to compare sensitivity and specificity between the two. The area under the receiver operating characteristics (AUROC) between NCEI + DWI and NEC + CEFST1 were compared by using DeLong Method [12]. All analyses were conducted using MedCalc for Windows (v. 20.1.1.1, MedCalc Software).
Interobserver agreements (κ) were calculated for NCEI + DWI and NCEI + CEFST1. The κ values were interpreted according to the following categories: poor (κ = 0), slight (κ = 0.0–0.2), fair (κ = 0.21–0.40), moderate (κ = 0.41–0.60), substantial (κ = 0.61–0.80), and almost perfect (κ = 0.81–1.00) [13].
RESULTS
The sensitivity and specificity of diagnosing infectious spondylitis were analyzed using NCEI + DWI and NCEI + CEFST1 in two readers. Results showed that the sensitivities and specificities of the two methods were not significantly different with 90.8% and 69.2% for NCEI + DWI, and 96.1% and 60.3% for NCEI + CEFST1 in reader 1, whereas 92.1% and 66.7% for NCEI + DWI, and 96.1% and 68.0% for NCEI + CEFST1 in reader 2, respectively (Table 2).
Table 2
Sensitivity, Specificity, and Accuracy for Infectious Spondylitis
Similarly, the sensitivities and specificities to diagnose abscess were 84.4% and 92.6% for NCEI + DWI, and 93.8% and 89.3% for NCEI + CEFST1 in reader 1, whereas 84.4% and 87.8% for NCEI + DWI, and 84.4% and 87.8% for NCEI + CEFST1 in reader 2. The analysis revealed that NCEI + DWI and NCEI + CEFST1 had comparable performance in both readers (Table 3).
Table 3
Sensitivity, Specificity, and Accuracy for Abscess
A significant difference was demonstrated in one reader when AUROC was compared between NCEI + DWI and NCEI + CEFST1 (Table 4). The representative cases for infectious spondylitis and Modic type 1 change were demonstrated in Figures 2 and 3, respectively. Interobserver agreement between the two readers was moderate (κ = 0.55) in NCEI + DWI and substantial (κ = 0.66) in NCEI + CEFST1.
Fig. 2
61-year-old male patient with back pain. A: Conventional T2-weighted image shows paraspinal high signal intensity (arrows) and fluid collection (arrowhead). B: Conventional T1-weighted image shows obliteration of paraspinal fat plane (arrows). C: Contrast-enhanced fat-suppressed T1-weighted image shows diffuse paraspinal enhancement (arrows) and abscess (arrowhead). D, E: These lesions (arrows) also show as diffuse high signal on diffusion-weighted imaging with b = 800 s/mm2 and intermediate to low signal on apparent diffusion-coefficient map. Corynebacterium striatum was cultured from the drain tip at the operation site, and these lesions were confirmed as infectious spondylitis after the operation.
Fig. 3
42-year-old male patient with fever and back pain. A, B: Conventional T2- and T1-weighted images show no signal change in and around the vertebral body. C: Contrast-enhanced fat-suppressed T1-weighted image shows high paraspinal signal, probably representing venous enhancement rather than paraspinal soft tissue enhancement (arrows). D, E: On diffusion-weighted imaging with b = 800 s/mm2 and apparent diffusion-coefficient map, there is no definite signal change in paraspinal soft tissue. This lesion was regarded as Modic type 1 endplate change with disc degeneration.
Table 4
Comparison of the Area Under Receiver-operating Characteristic Curves between Diffusion-weighted and Contrast-enhanced MRI
Of the three false-negative cases, two only showed subtle endplate enhancement on CEFST1, leading both readers to misidentify it as a Modic type 1 change. The third case showed para-spinal enhancement on CEFST1, but it was not obvious enough to be interpreted as infectious spondylitis on DWI (Fig. 4).
Fig. 4
64-year-old female patients with fever and back pain. A, B: Conventional T2- and T1-weighted images show subtle haziness in paraspinal fat (arrows). C: Contrast-enhanced fat-suppressed T1-weighted image shows paraspinal soft tissue enhancement (arrows), suggestive of infectious spondylitis. D, E: Diffusion-weighted imaging with b = 800 s/mm2 and apparent diffusion-coefficient map shows paraspinal soft tissue stripe (arrows), which were not confident enough to be interpreted as spondylitis.
DISCUSSION
The demand for MRI without contrast enhancement is growing due to the safety concerns of GD-based contrast media. However, the diagnostic performance of MRI without contrast enhancement should be maintained at an acceptable range compared to the contrast-enhanced MRI. We hypothesized that DWI has the potential to diagnostic aids in infectious spondylitis without contrast enhancement. To substantiate the hypothesis, we compared the diagnostic performance of non-enhanced conventional MRI with DWI and contrast-enhanced MRI for the diagnosis of infectious spondylitis.
Previous studies have revealed that the sensitivity of contrast-enhanced MRI is up to 90% or higher [14, 15, 16]. In comparison, our study showed that contrast-enhanced MRI and non-contrast-enhanced MRI with DWI had a similar sensitivity range in both readers. However, the specificity of both readers in our study was 60.3% and 68.0%, which was lower than the results of previous studies that indicated a specificity rate of over 90% [14, 15, 16]. Our study showed low specificity, probably due to the inclusion criteria of the control group. We included Modic type 1 and recently developed compression fractures with bone marrow signal alteration resembling spondylitis. In contrast, previous studies set the inclusion criteria based on the clinical suspicion of infectious spondylitis resulting in a higher specificity as more obvious negative cases on MRI were included. Although the conventional inclusion criteria align with the clinical scenario, we wanted to examine the ability to diagnose infectious spondylitis when there is a marrow signal change.
DWI has proven effective in diagnosing variable infectious conditions, including appendicitis, pelvic inflammatory disease, urinary tract infection, and intracranial infection [17, 18, 19, 20]. Studies have also demonstrated its potential for diagnosing infectious spondylitis [21, 22], with the results indicating that ADC values from the intervertebral disk area or high-signal intensity area on sagittal ADC map have a significantly different between infectious spondylitis and degenerative disk disease groups. Daghighi et al. [22] reported that sensitivity of 91.7% and specificity of 96.8% with a cut-off value of ADC. However, the quantitative approach may be subject to operator dependency and bias depending on the number of pixels included. DWI is generally acquired with fat suppression allowing bone marrow and paraspinal soft tissue edema to stand out against a suppressed background. In addition, diffusion gradient preferentially suppresses the signal from vessels and free water contained in degenerated or traumatic bone marrow [23, 24]. These combined effects allowed DWI to distinguish infected bone and soft tissue from the non-infected area. Moreover, we acquired axial DWI to include paraspinal soft tissue, which is vital for diagnosing infectious spondylitis. When comparing NCEI + CEFST1 with NCEI + DWI for diagnosing infectious spondylitis, there was no statistically significant difference in sensitivities and specificities in both readers. However, when comparing AUROC, the results were discrepant in both readers. Reader 2 showed a significantly higher AUROC for NCEI + CEFST1 than for NCEI + DWI, suggesting the low signal-to-noise ratio and low resolution of axial DWI compared to the axial-enhanced CEFST1 may lower the diagnostic confidence in less experienced readers.
The detection of the paraspinal abscess was evaluated as a secondary outcome. The presence of a paraspinal or epidural abscess often indicates infectious spondylitis. Previous studies have pointed out that DWI is useful for diagnosing soft tissue abscesses [25, 26]. DWI maintained a relatively high signal in high b-value images because it contains viscous internal contents of the abscess [27, 28]. Consequently, DWI was found to have more than 87% accuracy in both readers for detecting abscesses in our research. Moreover, the sensitivity and specificity of DWI for detecting abscesses associated with infectious spondylitis were comparable to that of contrast-enhanced MRI.
Previous reports have not compared the diagnostic performances of MRI with and without contrast material. Contrast-enhanced MRI is accepted as a standard of practice for delineating paraspinal and epidural abscesses, demonstrating the extent of the infected tissue, and distinguishing spondylitis from edema or tumors [29, 30]. However, signal change on both T1 and T2-weighted images in combination with the structural change of the endplates in conventional sequences without contrast enhancement may lead to the correct diagnosis of infectious spondylitis [29]. Axial fluid-sensitive sequences are also sensitive to paraspinal soft tissue edema and could be viable in cases where GD injection is contraindicated.
This study had several limitations. Microbiological confirmation was not made for all cases due to the difficulty of cultivating a minor strain of microorganisms. In addition, some of the patients received empirical antibiotic therapy before the MRI scan, so the diagnosis was based on laboratory and clinical features. Similarly, the control group was also enrolled without pathological confirmation. However, correlation with clinical features and follow-up is an ethically acceptable approach if the infectious spondylitis is not clinically suspected. Second, the criteria for the control group were somewhat arbitrary. However, we set the control group to simulate the diagnostic dilemma in imaging interpretation. Third, we excluded patients with metallic implants in the spine as DWI is non-diagnostic with the metallic artifact. However, it should be noted that infectious spondylitis can also occur in the presence of metallic implants meaning our results cannot be extrapolated to the postoperative patients with metallic instruments. Fourth, due to time constraints, DWIs were obtained with only 2 b-values even though previous studies recommend at least three b-values for increased precision of ADC map generation [25]. However, this recommendation was not followed since we visually assessed the DWI and ADC map without measuring ADC.
In conclusion, nonenhanced conventional MR images combined with axial DWI are a viable alternative to contrast-enhanced MRI for diagnosing infectious spondylitis in cases where the use of GD is not possible. This combination provides similar results to the use of contrast-enhanced MRI.
Conflicts of Interest:Joon-Yong Jung, a contributing editor of the Investigative Magnetic Resonance Imaging, was not involved in the editorial evaluation or decision to publish this article. All remaining authors have declared no conflicts of interest.
Author Contributions:
Conceptualization: Joon-Yong Jung.
Data curation: Jung-Min Hwang, So-Yeon Lee, Seung-Hoon Choi, Joon-Yong Jung.
Formal analysis: Jung-Min Hwang, Seung-Hoon Choi, Joon-Yong Jung.
Investigation: So-Yeon Lee, Joon-Yong Jung.
Methodology: Seungeun Lee, Joon-Yong Jung.
Project administration: Joon-Yong Jung.
Resources: Seung-Hoon Choi, Joon-Yong Jung.
Supervision: Joon-Yong Jung.
Validation: Seung-Hoon Choi, Seungeun Lee, Joon-Yong Jung.
Visualization: Seung-Hoon Choi, Joon-Yong Jung.
Writing—original draft: Seung-Hoon Choi, Joon-Yong Jung.
Writing—review & editing: Joon-Yong Jung.
Funding Statement:None
Availability of Data and Material
The datasets generated or analyzed during the study are available from the corresponding author on reasonable request.
References
-
Woolen SA, Shankar PR, Gagnier JJ, MacEachern MP, Singer L, Davenport MS. Risk of nephrogenic systemic fibrosis in patients with stage 4 or 5 chronic kidney disease receiving a group II gadolinium-based contrast agent: a systematic review and meta-analysis. JAMA Intern Med 2020;180:223–230.
-
-
Fleiss JL. Measuring nominal scale agreement among many raters. Psychol Bull 1971;76:378–382.
-
-
Simrén Y, Stokland E, Hansson S, Hebelka H, Svensson PA, Lagerstrand KM. Diffusion tensor imaging based multiparametric characterization of renal lesions in infants with urinary tract infections: an explorative study. BMC Pediatr 2021;21:440
-
-
Daghighi MH, Poureisa M, Safarpour M, et al. Diffusion-weighted magnetic resonance imaging in differentiating acute infectious spondylitis from degenerative Modic type 1 change; the role of b-value, apparent diffusion coefficient, claw sign and amorphous increased signal. Br J Radiol 2016;89:20150152
-
-
Martín Noguerol T, Luna Alcalá A, Beltrán LS, Gómez Cabrera M, Broncano Cabrero J, Vilanova JC. Advanced MR imaging techniques for differentiation of neuropathic arthropathy and osteomyelitis in the diabetic foot. Radiographics 2017;37:1161–1180.
-
-
Cartes-Zumelzu FW, Stavrou I, Castillo M, Eisenhuber E, Knosp E, Thurnher MM. Diffusion-weighted imaging in the assessment of brain abscesses therapy. AJNR Am J Neuroradiol 2004;25:1310–1317.
-
MeSH Terms
Figures
1 / 4
Tables
1 / 4
