If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Discriminating imaging features between bacterial and tuberculous pyomyositis.
•
T2 high signal intensity at abscess wall can predict for bacterial pyomyositis.
•
Abscess wall character on imaging of bacterial muscle infection.
•
Magnetic resonance imaging between bacterial and tuberculous muscle infection.
Abstract
Purpose
The purpose of this study was to assess the differences of magnetic resonance features between tuberculous and bacterial pyomyositis.
Method
This is a retrospective study of patients with bacterial and tuberculous pyomyositis. We excluded patients with pyomyositis caused by actinomycosis, non-tuberculous mycobacterium, fungi, unknown of causative organism, or inadequate imaging for analysis. Magnetic resonance imaging was independently reviewed by two radiologists.
Results
Of the 136 pyomyositis patients, 71 (52.2 %) patients had bacterial pyomyositis while 65 (47.8 %) patients had tuberculous pyomyositis. Seventy-seven patients (56.6 %) had intramuscular abscess. On multivariable analysis, bacterial pyomyositis was associated with diabetes mellitus (odds ratio [OR] 3.17, 95 % confidence interval [CI] 1.30–8.24) and bone marrow involvement (OR 5.02, 95 % CI 1.21–34.4). Spinal involvement had a significantly lower likelihood of bacterial pyomyositis (OR 0.25, 95 %CI 0.11–0.54). In patients with intramuscular abscess, diabetes mellitus and hyperintense on T2-weighted images at the abscess wall had a significantly higher likelihood of bacterial pyomyositis (OR 5.21, 95 %CI 1.33–25.42 and OR 5.34, 95 %CI 1.36–24.71, respectively), whereas spinal involvement had a significantly lower likelihood of bacterial pyomyositis (OR 0.09, 95 %CI 0.02–0.30).
Conclusions
Magnetic resonance imaging has modest accuracy for differentiation of tuberculous and bacterial pyomyositis. Diabetes mellitus and extraspinal pyomyositis were the predictors of bacterial pyomyositis. Presence of T2 hyperintense wall of intramuscular abscess was also the predictor of bacterial pyomyositis.
]. Most commonly, infectious pyomyositis is caused by bacteria, predominantly Staphylococcus aureus, whereas tuberculous and nonbacterial pyomyositis including virus, fungi and parasites are rare [
]. The etiology of pyomyositis is classified into primary and secondary pyomyositis. Primary pyomyositis is an infection of skeletal muscle itself, while secondary pyomyositis is a contiguous infection from the adjacent structures such as bone, joint or soft tissue [
Magnetic resonance imaging (MRI) is the most useful imaging modality for diagnosis of pyomyositis with high sensitivity and specificity. It most clearly demonstrates diffuse muscle inflammation as well as any subsequent abscess formation [
To reduce the risk of progression of pyomyositis to sepsis and mortality, it is important to differentiate tuberculous and bacterial pyomyositis for the proper treatment at the early stage of the disease. However, there is a paucity of literature discussing how to differentiate between these two causative organisms. This study aims to assess the magnetic resonance imaging features to differentiate tuberculous and bacterial pyomyositis.
2. Methods
2.1 Study design
2.1.1 Patient population
We retrospectively reviewed 355 patients who were diagnosed with pyomyositis or the following conditions of septic arthritis, osteomyelitis and spondylodiscitis and have undergone MRI between August 2013 and April 2019. We excluded the patients who had diagnosis of septic arthritis, osteomyelitis and spondylodiscitis with no muscle involvement in 55 patients. We excluded the patients who had chronic bacterial infection (Actinomyces, Mycobacterium abscessus, Mycobacterium avium complex) or fungal infection (Eumycetoma, Cryptococcus neoformans) in 8 patients, and inadequate imaging for analysis in 2 patients. One hundred fifty-four patients were also excluded because they did not have data documents about the causative organisms, and they were referred for only imaging studies. A total of 136 patients were included in Fig. 1. This study was approved (Approval number HE611541) by the Ethics Committee for Human Research.
] and by reviewing the tissue or fluid culture, hemoculture, polymerase chain reaction, Acid-Fast Bacilli stain, histopathology, the medical records of the causative organisms or based on responsiveness of antibiotics or antituberculous therapy.
2.1.2 Imaging protocol
MRI was performed on 1.5 T system (Magnetom Aera, Siemens Medical Solutions, Erlangen, Germany) or 3 T system (Philips Achieva, Philips Medical System, Netherland, B.V.). MRI protocol included a combination of axial, sagittal and /or coronal images using T1-weighted (T1WI) fast spin echo (FSE) with TR/TE range of 400–680/8–12, T2-weighted (T2WI) FSE with TR/TE range of 3000–5360/78–100 and Short Tau Inversion Recovery (STIR) or Spectral Attenuated Inversion Recovery (SPAIR); matrix size 256 × 160 to 384 × 216. Contrast enhanced images of axial, sagittal and/or coronal planes were obtained using T1WI FSE sequence with fat suppression after 0.1 mmol/kg of body weight of gadolinium injection. Field of view, slice thickness, and interslice gap varied along with diseased region.
2.1.3 Imaging analysis
The MR images were analyzed on Picture Archiving and Communication System (PACS). Two musculoskeletal radiologists, who had experiences about 10 and 20 years, were independently reviewed under blind condition to patient’s information and the diagnosis of causative organisms. The data were collected as double data entry. The discrepancy findings between radiologists were given in a consensus.
The extension of abnormalities was classified into two categories; a lesion in the isolated muscle alone or a muscle lesion with involvement of adjacent structures including bone marrow, joint, fistula, and spine. For more precise evaluation, the location of pyomyositis (lower extremity, upper extremity, neck, psoas muscle, pelvic iliacus muscle, paravertebral muscle, thoracic wall, buttock or others), the number of affected muscles (single or multiple muscles), muscle contour (normal or enlarged), characters of the affected muscles including signal intensity (hypointense, isointense, or hyperintense on T1-weighted images and T2-weighted images in comparison with normal skeletal muscle) and feathery muscle edema were taken in account. The presence of intramuscular gas was assessed in case of absent history of biopsy, drainage or fistula. The cellulitis was also evaluated.
Regarding to the presence of intramuscular abscess, which was defined by a mass with the signal intensity of fluid collection and a peripheral rim enhancement after intravenous gadolinium injection. We characterized the abscess in terms of the location (lower extremity, upper extremity, neck, psoas muscle, pelvic iliacus muscle, paravertebral muscle, thoracic wall, buttock or others), the number of abscess (single or multiple abscesses), the abscess volume [by the ellipsoid volume formula [4/3 x ¶ x (height/2) x (width/2) x (depth/2)] [
], signal intensity of the abscess wall (hypointense, isointense, or hyperintense on T1-weighted images and T2-weighted images in comparison with normal skeletal muscle), signal intensity of internal abscess content (hypointense, isointense, or hyperintense on T1-weighted images and T2-weighted images in comparison with normal skeletal muscle), the character of abscess wall (smooth or irregular wall was defined as less than or equal to or more than 3 mm, difference from each side of the abscess wall), the thickness of abscess wall (thick wall was defined as ≥ 3 mm thickness) and the presence of internal abscess enhancement. When there was more than one abscess, the largest abscess was evaluated.
Study data were collected and managed using Research Electronic Data Capture (REDCap) [
Research electronic data capture (REDCap)--a metadata-driven methodology and workflow process for providing translational research informatics support.
Continuous variables were presented as mean (standard deviation) or median (interquartile range [IQR], as appropriate. Categorical variables were reported as number (percentage). Continuous variables were compared between two groups (bacterial group vs tuberculosis group) using two sample t-test or Wilcoxon rank-sum as appropriate. Chi-square test or Fischer’s exact test were used for categorical variables.
Binary logistic regression was used for analyzing the predictors associated with bacterial pyomyositis. To build a multivariable regression model, univariable regression was first analyzed. The variables significant at p < 0.1 in univariable analysis were identified as potential factors variables and entered into a multivariable regression model. Four-fold cross validation was used in estimating the generalization error and model performance. Area under the receiver operating characteristic (ROC) were calculated from cross-validation for determination the model performance. We performed a sub-group analysis on patients who had the abscess then developed the predictive model for bacterial pyomyositis. All the statistical analyses were performed using R software version 3.5.2. The level of statistical significance was set at p < 0.05 (two tallied).
3. Results
A total of 136 patients were diagnosed as having pyomyositis. Of these patients, 71 patients (52.2 %) were bacterial pyomyositis and 65 patients (47.8 %) were tuberculous pyomyositis. Among 136 patients, 77 (56.6 %) of them had abscess. Thirty-seven (52.1 %) of 71 bacterial pyomyositis patients were males while 32 (49.2 %) of 65 tuberculous pyomyositis patients were males. The median age of both groups were not different (53 [IQR 43.5–61.5] vs 57 [42–65], p = 0.30). The pathogenic organisms detected by culture were Staphylococcus spp. in 14 patients (35 %), Streptococcus spp. in 9 patients (22.5 %), Burkholderia pseudomallei in 5 patients (12.5 %), Escherichia coli in 4 patients (10 %), Salmonella spp. in 3 patients (7.5 %), Pseudomonas aeruginosa in 2 patients (5 %), and 1 patient (2.5 %) each with Klebsiella pneumonia, Aeromonas veronii, and Acinetobacter baumannii. The patients with history of diabetes mellitus were more frequent in bacterial pyomyositis group than in tuberculous pyomyositis group (22 [31 %] vs 10 [15.4 %], p = 0.03) (Table 1).
Table 1Demographic Data between Pyogenic and Tuberculous Pyomyositis.
Characteristics
Tuberculous pyomyositis(n=65)
Pyogenic pyomyositis(n=71)
P-value
Age, median (IQR), years
57
(42-65)
53
(43.5-61.5)
0.30
Male, n (%)
32
(49.2)
37
(52.1)
0.74
Comorbidities, n (%)
Diabetes mellitus
10
(15.4)
22
(31)
0.03
Malignancy
2
(3.1)
8
(11.3)
0.10
Chronic kidney disease
5
(7.7)
5
(7)
1.00
Connective tissue disease
4
(6.2)
5
(7)
1.00
History of muscle trauma
1
(1.5)
5
(7)
0.21
HIV infection
2
(3.1)
1
(1.4)
0.61
Cirrhosis
2
(3.1)
1
(1.4)
0.61
Note—HIV = human immunodeficiency virus, IQR = interquartile range.
Pyomyositis at lower extremities were more frequent in bacterial pyomyositis group compared to tuberculous pyomyositis group (29 [40.8 %]) vs 10 [15.4 %], p < 0.001). The psoas muscle involvement was more common in tuberculous pyomyositis group compared to bacterial pyomyositis (39 [60 %] vs 29 [40.8 %]), p = 0.03). The majority of patients in both bacterial and tuberculous pyomyositis groups had involvement of muscle and other structures with predominant involvement of spines. Spinal involvement was more common in tuberculous pyomyositis than in bacterial pyomyositis (46/61 [75.4 %] vs 24/56 [42.9 %], p < 0.001) (Table 2). Other less frequently involved structures were the joint, bone marrow and fistula tract. Pyomyositis with bone marrow involvement was more frequent in patients with bacterial pyomyositis compared to tuberculous pyomyositis (15/56 [26.8 %] vs 2/61 [3.3 %], p < 0.001). An isolated muscle infection was more common in bacterial pyomyositis than in tuberculous pyomyositis (15 [21.1 %] vs 4 [6.2 %], p = 0.01).
Table 2Magnetic Resonance Imaging Findings of Pyogenic and Tuberculous Pyomyositis in All Patients.
3.2 MRI findings of pyomyositis with intramuscular abscess
Among the patients who have pyomyositis with abscess formation, involvement of lower extremities was significantly more frequent in bacterial pyomyositis group than in tuberculous pyomyositis group (16 [40 %] vs 6 [16.2 %], p = 0.02). Similarly, involvement of buttock was more common in patients with bacterial pyomyositis with abscess compared to tuberculous pyomyositis with abscess (23 [57.5 %] vs 13 [35.1 %], p = 0.05). Isolated muscle involvement was also significantly more common in bacterial than tuberculous pyomyositis with abscess (11 [27.5 %] vs 3 [8.1 %], p = 0.03). Bone marrow involvement was significantly higher in bacterial pyomyositis with abscess compared to that in tuberculous pyomyositis with abscess (7/29 [24.1 %] vs 1/34 [2.9 %], p = 0.02). In contrast, patients with spine involvement were more common in tuberculous pyomyositis with abscess, compared to those in bacterial pyomyositis with abscess (27/34 [79.4 %] vs 10/29 [34.5 %], p < 0.001) (Table 3).
Table 3Magnetic Resonance Imaging Findings of Pyogenic and Tuberculous Pyomyositis in Patients with Intramuscular Abscess.
On multivariable analysis, bacterial pyomyositis showed significant association with the history of diabetes mellitus (odds ratio [OR] 3.17, 95 % confidence interval [CI] 1.30–8.24, p = 0.01) and bone marrow involvement (OR 5.02, 95 %CI 1.21–34.4, p = 0.05), although spinal involvement had a significantly lower association with bacterial pyomyositis (OR 0.25, 95 %CI 0.21–0.54, p < 0.001) (Table 4). The area under the receiver operating characteristic curve (95 %CI) of this multivariable analysis was 0.7 (95 %CI 0.61–0.79).
Table 4Factors Associated with Pyogenic Pyomyositis.
On multivariable analysis for the sub-group of patients with intramuscular abscess, a history of diabetes mellitus had a significantly higher likelihood of bacterial pyomyositis (OR 5.21, 95 %CI 1.33–25.42, p = 0.03) whereas spinal involvement had a significantly lower likelihood of bacterial pyomyositis (OR 0.09, 95 %CI 0.02–0.30, p < 0.001). In addition, hyperintensity on T2-weighted image at the abscess wall was a significant predictor of bacterial pyomyositis (OR 5.32, 95 %CI 1.36–24.71, p = 0.02) (Table 4). The area under the receiver operating characteristic curve (95 %CI) of this multivariable analysis was 0.77 (95 %CI 0.66–0.88).
4. Discussion
Differentiation between bacterial and tuberculous pyomyositis is crucial. Our study revealed that patients with a history of diabetes mellitus were strongly associated with bacterial pyomyositis. Site and extension of infection from MR imaging were important for differential diagnosis between bacterial and tuberculous pyomyositis. Patients with bone marrow involvement were associated with bacterial pyomyositis while patients with spine involvement were associated with tuberculous pyomyositis. Among patients who developed intramuscular abscess, hyperintensity of abscess wall on T2-weighed images was associated with bacterial pyomyositis.
There are many comorbidities that might impair the immune system of patients with pyomyositis such as diabetes mellitus, human immunodeficiency virus infection, liver cirrhosis, malignancy, autoimmune disease, connective tissue disease or chronic kidney disease [
]. In our study, diabetes mellitus was the significant predictive factor for development of bacterial pyomyositis. This supported the previous studies that diabetes mellitus was a predisposing factor for pyogenic spondylodiskitis [
In the previous series of MRI study, characteristic features of pyomyositis were described as muscle enlargement, with increased signal intensity on T2-weighed images, usually associated with subcutaneous fat stranding and edema along the fascial planes [
]. Similar to our patients, all cases showed increased signal intensity on the T2-weighed images at the muscles and most of cases had enlarged muscle contours, and there were no differences between bacterial and tuberculous pyomyositis groups.
On T1-weighted images, the intramuscular abscess is often characterized by hyperintense which was found in both bacterial and tuberculous pyomyositis in our study. This finding was consistent with the others [
] and it is postulated that hyperintense rim could be related to the presence of paramagnetic materials in the abscess wall such as blood product, bacterial or macrophage sequestration of iron or free radicals. In addition, in this study we have found that hyperintense rim on T2-weighted images of the intramuscular abscess wall was a significant predictor of bacterial pyomyositis as shown in Fig. 2. In the previous report of Stavroula et al. [
] reported that the abscess wall was hypointense on T2-weighted images in tuberculous pyomyositis. These findings in the previous studies are not conclusive. We assume that these findings are related to the age of blood product and vascularized granulation tissue in the abscess wall as shown in Fig. 3 here. According to the nature of disease, tuberculosis is more chronic than bacterial infection. In bacterial pyomyositis, earlier stage of blood production, more vascularized granulation tissue and less fibrotic tissue component give rise MR imaging of hyperintense rim on T2-weighted images at the abscess wall.
Fig. 263-year-old man with diabetic mellitus and primary pyogenic pyomyositis from beta-hemolytic streptococcus group A. Axial T2-weighted fast spin echo (FSE) image (A) shows muscle enlargement and edema at the arm. There are discrete abscesses in the biceps and triceps muscles. The abscess wall shows T2 hyperintense rim, compared to adjacent normal skeletal muscle (white arrows). Axial contrast-enhanced T1-weighted FSE image (B) shows smooth thin-walled abscess without internal septal enhancement (black arrow).
]. Similar to our results, the lower extremities muscles were the one of the most commonly affected location in bacterial pyomyositis. In addition, muscular lesion with bone marrow involvement was significantly more common in bacterial pyomyositis group than in tuberculous group. In contrast, patients with spine involvement were associated with a lower likelihood of bacterial pyomyositis as shown in Fig. 4. To our knowledge, there is no previous report mentioned statistically about this information.
Fig. 439-year-old man with L3/4 tuberculous spondylodiscitis and secondary pyomyositis at right psoas muscle. Sagittal T1-weighted, T2-weighted and contrast-enhanced T1-weighted fast spin echo (FSE) images (A)(B)(C) show L3/4 intervertebral disc and adjacent bony destruction. Coronal fat-suppressed T2-weighted FSE image (D) and axial T2-weighted FSE image (F) show right psoas muscle enlargement with intramuscular abscess and the abscess wall shows hypointensity (white arrows). Axial T1-weighted FSE image (E), the abscess wall appears hypointensity (black arrows) and on axial contrast-enhanced T1-weighted FSE image (G), the abscess wall appears smooth and thin.
In the present study, data quality was assured by double data entry and was given consensus from two radiologists. In addition, we used the modern machine learning techniques for estimating the model performances. However, there are limitations due to small sample size of study populations. Moreover, we included patients who responded to antimicrobial therapy alone, so that severe cases may have been unintentionally excluded if their treatment failed. Future study should evaluate the relationship between MRI findings and histopathological findings in patients with intramuscular abscess.
5. Conclusion
MRI has modest accuracy for differentiation of bacterial and tuberculous pyomyositis. Diabetes mellitus and extraspinal pyomyositis were the predictors of bacterial pyomyositis with and without intramuscular abscess, compared to those of tuberculous pyomyositis. Presence of T2 hyperintense wall of intramuscular abscess was the predictor of bacterial pyomyositis.
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Declaration of Competing Interest
None.
Acknowledgments
We would like to acknowledge Professor Yukifumi Nawa for editing the manuscript via Publication Clinic of Khon Kaen University, Thailand.
References
Soler R.
Rodriguez E.
Aguilera C.
Fernandez R.
Magnetic resonance imaging of pyomyositis in 43 cases.
Research electronic data capture (REDCap)--a metadata-driven methodology and workflow process for providing translational research informatics support.
30003-4&id=gr1.jpg){kind=link}
30003-4&id=gr2.jpg){kind=link}
30003-4&id=gr3.jpg){kind=link}
30003-4&id=gr4.jpg){kind=link}