| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Clinical Observations |
1 All authors: Department of Radiology, Mayo Clinic, 200 First St. SW, Charlton 2-290, Rochester, MN 55905.
Received July 28, 2004; accepted after revision October 1, 2004.
Address correspondence to M. S. Collins
(collins.mark{at}mayo.edu).
OBJECTIVE. The objective of our study was to better define the T1-weighted MRI characteristics of surgically proven pedal osteomyelitis.
CONCLUSION. Decreased T1 marrow signal in a geographic medullary distribution with a confluent pattern and concordance with fat-suppressed T2- and T1-weighted postcontrast signal abnormality was present in 100% of the surgically proven cases of pedal osteomyelitis. None of the patients with decreased T1 marrow signal in a subcortical distribution or in a hazy, reticulated pattern had surgically proven osteomyelitis regardless of the fat-suppressed T2-weighted or postcontrast T1-weighted findings.
Prompt clinical recognition of pedal osteomyelitis is important because early aggressive medical and surgical interventions are imperative for adequate patient management. Imaging options include radiography, nuclear medicine studies, and MRI. Radiographs are important for initial patient screening and are helpful when cortical bone abnormalities are present; however, they have low sensitivity. Nuclear medicine studies have improved sensitivity compared with radiographs but have low specificity and are less helpful than MRI in evaluating adjacent soft-tissue abnormalities. MRI has been shown to be a useful imaging study that is capable of revealing primary marrow abnormalities and secondary bone and soft-tissue abnormalities with improved specificity as compared with scintigraphy in patients with clinically suspected pedal osteomyelitis.
Prior studies have identified reliable primary and secondary MRI features that suggest osteomyelitis [1–14]. Primary MRI findings of osteomyelitis include bone marrow signal abnormality consisting of decreased T1 signal, increased T2 signal, and postgadolinium enhancement. Supportive secondary MRI findings of osteomyelitis include adjacent soft-tissue ulcers, cellulitis, phlegmon, abscess, sinus tracts, and cortical bone destruction [1]. Prior studies have emphasized abnormal findings on T2-weighted and gadolinium-enhanced imaging sequences. The primary MRI findings of osteomyelitis (i.e., decreased T1 signal and increased T2 marrow signal) also may be seen in other pedal conditions including reactive marrow edema, stress reaction, and neuropathic arthropathy, which decreases MRI specificity. Abnormal T2 signal also may be seen artifactually in areas of inhomogeneous fat saturation that may occur at curved surfaces such as the heel and forefoot. In addition, postgadolinium marrow enhancement may be seen in areas of reactive marrow edema [3].
A clear distinction of normal and abnormal bone is critical for surgical planning in cases that fail to respond to medical management and require surgical débridement or amputation. Our clinical observation has been that fat-suppressed T2-weighted sequences interpreted independently of corresponding T1-weighted sequences may be overly sensitive in accurately delineating areas of true osteomyelitis (Fig. 1A, 1B, 1C, 1D). The purpose of this study was to directly evaluate the reliability of primary T1 findings in surgically proven cases of pedal osteomyelitis.
|
|
|
|
All patients from our institution with an MRI diagnosis of pedal osteomyelitis over a 6-year period (1996–2002) were identified using a computer diagnosis search engine. Those patients who went on to tissue diagnosis with surgical biopsy or amputation constituted our study group, which consisted of 80 feet in 80 patients who ranged in age from 22 to 89 years (mean age, 62 years). Sixty-three (79%) of these patients have diabetes.
The study group was divided into two groups: true-positive cases (MRI originally interpreted as osteomyelitis and surgical pathology interpreted as osteomyelitis) or false-positive cases (MRI originally interpreted as osteomyelitis and surgical pathology interpreted as negative for osteomyelitis). A positive histopathologic diagnosis of osteomyelitis was established on the basis of the presence of inflammatory cells, including neutrophils, lymphocytes, histiocytes, and plasma cells, within the marrow cavity of the submitted surgical specimen.
No statistical difference between the two groups was detected with regard to patient age, sex, presence of diabetes, or anatomic distribution of the affected bone. The study protocol was reviewed by the institutional review board who approved a retrospective review of the medical charts and diagnostic imaging of all the study patients.
MRI was performed with a 1.5-T superconducting magnet (Signa, GE Healthcare). A dedicated foot and ankle extremity coil was used in all cases with a field of view of 12–18 cm and slice thickness of 3 mm. Images were obtained in at least two orthogonal planes. In most cases, images were obtained in three planes. T1-weighted conventional spin-echo images (TR/TE range, 500/10–20) were obtained with 1–2 signals and a matrix size of 256 x 256. Fat-suppressed T2-weighted fast spin-echo images (TR range/TEeff range, 3,500–4,000/70–90) were obtained with 2 signals, an echo-train length of 8, and a matrix of 256 x 256. In a few cases, fast spin-echo inversion recovery sequences (TR range/TE, 3,000–5,000/50; inversion time, 150 msec; echo-train length, 8; 2 signals; and matrix, 256 x 192) were used in place of the T2-weighted fast spin-echo sequences. Fat-suppressed gadolinium-enhanced T1-weighted spin-echo sequences were obtained in 16 patients (10 of the true-positive group and six of the false-positive group) approximately 2 min after IV administration of 0.1 mmol/kg of gadolinium. Fat suppression was obtained in the T2-weighted fast spin-echo and gadolinium-enhanced T1-weighted spin-echo sequences using selective presaturation of lipid resonance peaks.
The MR images of all the study cases were reviewed retrospectively by two experienced musculoskeletal radiologists with a consensus interpretation. The reviewers were aware of the histopathologic diagnosis at the time of their analysis. Primary T1 imaging characteristics evaluated included T1 signal intensity of the affected bone marrow as compared with normal adjacent fatty marrow (decreased vs normal), T1 signal intensity as compared with adjacent skeletal muscle (decreased or isointense vs increased), the distribution of the signal abnormality (subcortical vs medullary), and the pattern (hazy, reticulated vs confluent).
T1 marrow distribution was defined as medullary if it involved a geographic area of the marrow canal. The distribution was considered subcortical if it involved a thin linear region of abnormal signal immediately adjacent to the cortical bone that typically measured no more than a few millimeters in depth. The T1 pattern was considered confluent if the marrow fat was completely replaced by a geographic area of decreased T1 signal. The T1 pattern was defined as hazy, reticulated if the marrow fat was not completely replaced, which was shown as areas of interspersed fatty signal throughout the area of hazy, decreased T1 signal abnormality (Figs. 2A, 2B and 3A, 3B).
|
|
|
|
Primary marrow findings were also assessed on T2-weighted fast spin-echo and fast spin-echo inversion recovery images. The degree of increased T2 signal abnormality was characterized as showing signal that either approached that of joint fluid (++) or was less than that of joint fluid (+). T1- and T2-weighted images were considered concordant if the marrow had geographic areas of T2 signal intensity that approached (++) or was less than (+) that of joint fluid that were associated with decreased T1 signal in a matching geographic medullary distribution with a confluent pattern. The T1- and T2-weighted sequences were considered discordant if the geographic areas of increased T2 signal were normal or had a hazy, reticulated pattern on the T1-weighted images (Fig. 4A, 4B).
|
|
All analyzed data were compared between the true-positive (positive MRI findings and positive pathology results) and false-positive (positive MRI findings and negative pathology results) groups using Fisher's exact test. Characteristics were considered statistically significant for p values less than 0.05. Positive predictive values were calculated for some of the individual characteristics.
Results
Of the 80 patients in the study group, findings were true-positive for 59 and false-positive for 21. Patients ranged in age from 22 to 89 years with a mean age of 62 years. Seventy-nine percent of the 80 patients had diabetes. The true-positive and false-positive groups were comparable: No statistical difference with regard to patient age, sex, or the presence of diabetes was detected. The time interval between the MRI examination and subsequent surgery was less than 6 weeks in all cases. The anatomic distribution of the 59 true-positive cases were phalanges, 25; metatarsals, 24; calcaneus, seven; cuboid bone, one; navicular bone, one; and cuneiform bone, one. The anatomic distribution of the 21 false-positive cases were metatarsals, 11; phalanges, seven; calcaneus, one; navicular bone, one; and talus, one.
With respect to the primary T1 findings, all patients in the true-positive group had decreased T1 marrow signal compared with normal adjacent fatty marrow. The abnormal area of decreased T1 signal was iso- or hypointense compared with muscle in 59 (100%) of 59 cases. This area of decreased T1 signal was located in a geographic medullary distribution and showed a confluent pattern in all of the true-positive cases (59/59, 100%). In the false-positive group, 19 (90.5%) of 21 had decreased T1 marrow signal compared with normal adjacent fatty marrow that was iso- or hypointense compared with muscle. In two of the 21 false-positive group cases, the T1 marrow signal was normal compared with the normal adjacent fatty marrow. The area of decreased T1 signal in the false-positive group was located in a geographic medullary distribution in 15 (78.9%) of 19 and in a subcortical distribution in four (21.1%) of 19 cases. The area of decreased T1 signal in the false-positive group showed a confluent pattern in eight (42.1%) of 19 and a hazy, reticulated pattern in 11 (57.9%) of 19 cases.
Statistical comparison of the primary T1 marrow abnormalities in the true-positive and false-positive groups using Fisher's exact test showed that both geographic medullary distribution and confluent pattern were statistically significant (p < 0.05). The calculated positive predictive value of decreased T1 marrow signal with geographic medullary distribution was 80%, and the positive predictive value for decreased T1 marrow signal with confluent pattern was 88%. No patients with surgically proven osteomyelitis had decreased T1 marrow signal with a subcortical distribution or hazy, reticulated pattern.
All study cases showed increased T2 marrow signal. The primary T2 marrow signal abnormality in the true-positive group approached the signal intensity of joint fluid (i.e., ++) in 54 (91.5%) of 59 cases compared with 13 (61.9%) of 21 cases in the false-positive group. The T2 signal abnormality was less bright than the signal intensity of joint fluid (+) in five (8.5%) of the 59 patients in the true-positive group versus eight (38.1%) of the 21 patients in the false-positive group. The positive predictive value of the finding that the T2 marrow signal abnormality approached the signal intensity of joint fluid (++) was 80% versus 38% for the group with a T2 signal abnormality that was less bright than the signal intensity of joint fluid (+).
The primary T1 and T2 marrow signal abnormality was considered concordant (i.e., decreased T1 signal with confluent pattern and a matching area of increased T2 signal intensity) in 59 (100%) of the 59 true-positive cases versus eight (38.1%) of the 21 false-positive cases. The marrow signal was discordant (i.e., T1 marrow signal normal or hazy, reticulated pattern with increased T2 signal abnormality) in zero (0%) of the 59 true-positive cases versus 13 (61.9%) of the 21 false-positive cases. The calculated positive predictive value for concordance of the T1 and T2 marrow abnormality was 88% (Table 1).
|
In 16 of the 80 study patients (10 of the true-positive group and six of the false-positive group), fat-suppressed gadolinium-enhanced T1-weighted spin-echo images were obtained. The primary marrow abnormalities of these 16 patients essentially paralleled the T2 findings in regard to geographic areas of increased T2 signal matching medullary enhancement and the concordance of the sequences with the unenhanced T1-weighted images. Therefore, no additional information was obtained with regard to the primary marrow findings, except one case with surgically proven intraosseous abscess of the calcaneus that showed areas of nonenhancing fluid surrounded by an area of rim enhancement; this finding was best shown by the fat-suppressed gadolinium-enhanced T1-weighted spin-echo sequences. The marrow space surrounding the abscess showed the typical features of decreased T1 and increased T2 signal intensity and postgadolinium enhancement. All the secondary soft-tissue findings, including soft-tissue ulceration, sinus tracts, and abscess formation, were subjectively better evaluated on the fat-suppressed gadolinium-enhanced T1-weighted spin-echo images.
|
Discussion
Prompt diagnosis of clinically suspected pedal osteomyelitis is essential to initiate early medical and surgical therapy. Diagnostic imaging options used for the evaluation of pedal osteomyelitis include various combinations of conventional radiographs, nuclear medicine studies, and MR images. MRI has been shown to be the most useful imaging study because it is capable of reliably detecting primary marrow signal abnormalities and secondary bone and soft-tissue abnormalities including cortical bone destruction, cellulitis, soft-tissue ulceration, phlegmon, abscess, and sinus tracts. Some patients may be successfully treated with early antibiotic therapy. However, many patients, especially those with diabetes, may require aggressive surgical débridement or selective amputation. Although highly effective, surgical treatment and anesthesia increase the cost of care and may be associated with higher morbidity and mortality, further emphasizing the necessity of an accurate imaging diagnosis.
|
Prior imaging studies have emphasized abnormal signal within the bone marrow on T2- and gadolinium-enhanced T1-weighted sequences. Morrison et al. [2] showed a sensitivity of 88% and specificity of 93% for fat-suppressed contrast-enhanced MR sequences versus 79% sensitivity and 53% specificity for unenhanced MRI in 51 cases of suspected osteomyelitis. However, Craig et al. [5] showed that postcontrast sequences were of limited use and were unable to distinguish areas of true osteomyelitis from areas of reactive edema. Morrison et al. [3] also emphasized the utility of secondary soft-tissue findings that may increase the sensitivity and specificity of the MR diagnosis. They showed that the secondary findings of cortical bone interruption, cutaneous ulcer, and a sinus tract adjacent to areas of marrow signal abnormality had the highest positive predictive value in identifying osteomyelitis in their study group, which included 43 patients with osteomyelitis. Thirty-four of the 43 patients in their study had histologic analysis of the marrow for definitive diagnosis of osteomyelitis. All 80 of our study patients had histologic confirmation of their diagnosis. Morrison et al. [3] also concluded that the primary signs of abnormal marrow signal intensity can be equivocal, and the findings on T1-weighted or fat-suppressed gadolinium-enhanced T1-weighted spin-echo images appear more reliable and reproducible than those on fat-suppressed T2-weighted or fast spin-echo inversion recovery images.
Schmid et al. [9] evaluated the performance of STIR versus fat-suppressed gadolinium-enhanced T1-weighted spin-echo images in 51 patients with a variety of bone marrow abnormalities of the foot and ankle. They found no difference between the two groups in the patterns of marrow signal abnormality [9]. Because contrast-enhanced images add to the time needed for imaging and the overall cost of the examination, they preferred to use STIR sequences. We primarily use fat-suppressed contrast-enhanced T1-weighted sequences to better define soft-tissue abnormalities associated with osteomyelitis rather than to further evaluate the marrow signal abnormality. STIR and fast spin-echo inversion recovery images are fluid-sensitive and can occasionally overstate the degree of marrow involvement [9].
The strengths of our study include the fact that all evaluated true-positive and false-positive cases had surgically obtained histopathologic confirmation of the final diagnosis. This eliminated the inclusion of presumed cases of osteomyelitis treated with antibiotics that may have truly represented reactive marrow edema or other causes of false-positive MRI abnormalities. Also, we specifically defined and identified primary T1 signal abnormalities that have not, to our knowledge, been previously specified (i.e., geographic medullary distribution and confluent pattern). We would like to emphasize that in our study cases the presence of ill-defined, decreased T1 signal within the marrow with a hazy, reticulated pattern (as opposed to a confluent pattern) is strongly suggestive of reactive marrow edema regardless of the T2-weighted or contrast-enhanced imaging findings (Fig. 6).
We found that the fat-suppressed T2-weighted sequences were less reliable in the ability to distinguish between the true-positive and false-positive groups. Also, a subcortical (rather than medullary) distribution of the T1 signal abnormality was seen only in the false-positive group and was usually attributed to reactive marrow edema. This was commonly seen in areas adjacent to soft-tissue ulcers and was sometimes seen in cases of completely exposed bone. All 59 patients in the true-positive group and 16 (76.2%) of the 21 patients in the false-positive group had abnormalities of the bone cortex immediately overlying the region of marrow signal abnormality. Seven (33%) of the patients in the false-positive group had complete disruption of the bone cortex that also occurred adjacent to areas of soft-tissue ulceration with completely exposed bone.
The limitations of our study included the retrospective design and lack of true- and false-negative cases because of our method of obtaining cases. This prevented us from being able to calculate test sensitivity and specificity. By excluding cases that were managed nonoperatively with antibiotic therapy, we perhaps selected a more complicated subset of cases. However, this is countered by the fact that surgical tissue diagnosis is considered the gold standard method of diagnosing osteomyelitis for study purposes. The cases were interpreted in consensus by two reviewers who were not blinded to the histopathologic results.
It would be helpful to conduct a prospective blinded study applying our statistically significant criteria. Unfortunately, there was overlap in the abnormal findings between the true- and false-positive groups. In the false-positive group, 78% had decreased T1 marrow signal with a medullary distribution, but only 42% of this group had a confluent pattern. We did not specifically distinguish clinically between cases of acute versus chronic osteomyelitis. Also, we did not specifically identify cases having coexisting osteomyelitis and septic arthritis. None of our study cases were diagnosed as neuropathic arthropathy at the initial MR interpretation or at the final histopathologic diagnosis.
When a patient is referred to MRI for evaluation of pedal osteomyelitis, it is important to know the exact site of clinically suspected osteomyelitis, typically centered on an area of overlying soft-tissue ulceration, to perform a targeted high-resolution study. High-resolution T1-weighted sequences should be performed in all three anatomic planes. Our current imaging protocol uses postcontrast high-resolution T1-weighted fast spin-echo images (TR/TE, 500/minimum full) with an echo-train length of 4, 2 signals, and a matrix of 256 x 256 interpolated to 512 x 256. Fast spin-echo inversion recovery images are obtained in two or three planes as directed on the basis of the T1 findings. Gadolinium-enhanced T1-weighted fast gradient-recalled echo sequences with fat saturation are performed in selected cases, especially to better define soft-tissue abnormalities.
When analyzing MR images of suspected cases of pedal osteomyelitis, it is helpful to begin with the fast spin-echo inversion recovery sequences to initially identify the generalized areas of marrow and soft-tissue signal abnormality. If available, gadolinium-enhanced T1-weighted sequences are helpful for confirming the fast spin-echo inversion recovery findings and more carefully scrutinizing the soft-tissue abnormalities, such as distinguishing between areas of cellulitis or phlegmon versus a defined abscess that requires more immediate clinical attention. Definitively, the unenhanced high-resolution T1-weighted fast spin-echo images are carefully evaluated in all three anatomic planes. If decreased marrow signal on the T1-weighted fast spin-echo images is not present in a geographic medullary distribution with a confluent pattern, we exclude the diagnosis of pedal osteomyelitis and attribute marrow abnormalities detected on fast spin-echo inversion recovery images as likely due to reactive marrow edema.
In conclusion, our study, which emphasized imaging features on T1-weighted images in surgically proven cases of pedal osteomyelitis, showed that direct primary findings of decreased T1 marrow signal in a geographic medullary distribution with a confluent pattern and concordance with the T2 marrow signal abnormality were present in 100% of true-positive cases. No cases of surgically confirmed osteomyelitis had primary T1 findings with subcortical distribution; hazy, reticulated pattern; or discordance with the T2-weighted images.
Imaging features that were statistically significant (p < 0.05) in distinguishing the true-positive group from the false-positive group included decreased T1 signal in a geographic medullary distribution, confluent pattern, concordant T1 and T2 signal abnormalities, and abnormalities in the adjacent bone cortex. We recommend that when evaluating the MR images of cases with clinically suspected pedal osteomyelitis, emphasis should be placed on meticulous assessment of marrow signal abnormality on triplanar high-resolution T1-weighted images. Fat-suppressed T2-weighted and gadolinium-enhanced sequences should be interpreted in conjunction with the T1 findings we have described.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  P. W. Johnson, M. S. Collins, and D. E. Wenger Diagnostic Utility of T1-Weighted MRI Characteristics in Evaluation of Osteomyelitis of the Foot Am. J. Roentgenol., January 1, 2009; 192(1): 96 - 100. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
