Original Research JOURNAL OF MAGNETIC RESONANCE IMAGING 22: (2005)

Transcription

1 JOURNAL OF MAGNETIC RESONANCE IMAGING 22: (2005) Original Research STIR vs. T1-Weighted Fat-Suppressed Gadolinium- Enhanced MRI of Bone Marrow Edema of the Knee: Computer-Assisted Quantitative Comparison and Influence of Injected Contrast Media Volume and Acquisition Parameters Marius E. Mayerhoefer, MD, 1 * Martin J. Breitenseher, MD, 1,2 Josef Kramer, MD, 3 Nicolas Aigner, MD, 4 Cornelia Norden, MD, 5 and Siegfried Hofmann, MD 6 Purpose: To compare short tau inversion recovery (STIR) and T1-weighted (T1w) gadolinium (Gd)-enhanced fat-suppressed MRI of bone marrow edema (BME) of the knee, and investigate the influence of injected contrast media volume and variation of major acquisition parameters on apparent BME volume and signal contrast. Materials and Methods: STIR and T1w Gd-enhanced fatsuppressed images were obtained from 30 patients with BME of the knee. Two groups of patients were examined with different MR scanners, acquisition parameters, and contrast media volumes. For both sequences, BME volume and signal contrast were assessed by computer-assisted quantification, and were compared through their arithmetic means and correlation coefficients (r 2 ). The injected contrast media volume was also correlated with BME volume and signal contrast differences between sequences. Results: A strong correlation between the STIR and Gdenhanced T1w images was found for BME volume (r ) and BME signal contrast (r ). Despite the differences in MR acquisition parameters and injected contrast media volume, both sequences depicted an almost identical BME volume in both groups. Contrast media volume showed a moderate correlation (r ) with BME volume differences. Conclusion: STIR is the optimum method for determining the size and signal contrast of BME. The injected contrast 1 Department of Radiology, Osteoradiology and Surgical Radiology Section, Medical University of Vienna, Vienna, Austria. 2 Waldviertelklinikum Horn, Horn, Austria. 3 Institute of CT and MRI Diagnostics Schillerpark, Linz, Austria. 4 First Orthopaedic Department, Orthopaedic Hospital Vienna-Speising, Vienna, Austria. 5 Schering AG, Clinical Development Cardiovascular, Berlin, Germany. 6 Department of Orthopaedics, LKH Stolzalpe, Stolzalpe, Austria. *Address reprint requests to: M.E.M., Department of Radiology, E7, Vienna General Hospital (AKH) Waehringer Guertel 18-20, 1090 Vienna, Austria. marius_mayerhoefer@aon.at Received January 18, 2005; Accepted August 9, DOI /jmri Published online 3 November 2005 in Wiley InterScience (www. interscience.wiley.com). media volume appears to have only a limited influence on apparent BME volume. Key Words: knee, MR; bone marrow; abnormalities; magnetic resonance (MR); inversion recovery; contrast enhancement; volume measurement J. Magn. Reson. Imaging 2005;22: Wiley-Liss, Inc. THE USE OF INTRAVENOUS contrast media injection for MRI of bone marrow disorders has long been the subject of discussion. While injection of contrast media is vital for MRI of neoplastic processes, its value for nonspecific lesions, such as bone marrow edema (BME), is still controversial. This is mainly because native MR sequences, such as short tau inversion recovery (STIR), also allow sensitive evaluation of bone marrow signal abnormalities. Although the underlying cause for abnormal signal intensity differs for contrast media-enhanced and native MR sequences, the apparent imaging pattern of BME is often almost identical (1), with ill-defined borders and a gradual transition of signal between normal and edematous tissue. It was the goal of this study to compare the volume and signal intensity of BME of the knee joint depicted by STIR and T1-weighted (T1w) contrast-enhanced fat-suppressed MR images, which are frequently used for MR examinations of bone marrow. This was done to determine whether the results were consistent with those from a similar study of bone marrow abnormalities of the foot and ankle, in which no significant difference between STIR images and gadolinium (Gd)- enhanced images with regard to the apparent BME volume was found (1). In addition, the influence of variations of injected contrast media volume and major MR acquisition parameters on the differences in apparent BME volume and signal contrast between the two MR sequences was also investigated Wiley-Liss, Inc. 788

2 STIR vs. Gadolinium-Enhanced MRI of BME 789 MATERIALS AND METHODS Patients and Imaging This study included 30 patients (eight women and 22 men, age range years, mean age 55 years) with BME of the knee (at least one bone affected) depicted on coronal turbo spin-echo (SE) STIR MR images. The study was approved by the institutional review board of two hospitals, and written informed consent was obtained from each patient. Based on the patients clinical history and findings on the STIR MR images, the following reasons for the development of BME were identified in the patient population: earlystage osteonecrosis (N 16), osteoarthritis (N 8), bone bruise (N 3), and stress fracture (N 3). For all patients, coronal T1w Gd-enhanced (gadopentetate dimeglumine, Magnevist; Schering, Berlin, Germany) fat-suppressed SE MR images were obtained in addition to the STIR images. Three months (range days; mean 87 days) after the initial MR examinations, all patients but one were examined a second time, for a total of 59 MR examinations. Because distal femoral condyles and proximal tibial condyles were regarded as two separate entities, a total of 118 study objects were available for image analysis. The MR images were obtained at two participating centers, and thus there were two different patient groups. Group 1 consisted of 16 patients (three women and 13 men, age range years; mean age 52 years; 64 study objects) who were examined with a 1.0 Tesla MR scanner (Philips/Marconi, Best, The Netherlands) using a dedicated knee coil. For the coronal turbo SE STIR images, a repetition time (TR) of 2500 msec, echo time (TE) of 10 msec, inversion time (TI) of 100 msec, and flip angle of 90 were used. For the coronal T1w Gd-enhanced fat-suppressed SE images, a TR of msec, TE of 18 msec, and flip angle of 90 were used, with an intravenous contrast media bolus injection of 0.2 ml ( 0.1 mmol) per kilogram body mass. For both sequences, a slice thickness of 4 mm and a acquisition matrix were used. Images were stored in DICOM format, with a grayscale depth of 10 bits (1024 grayscale values). Group 2 consisted of 14 patients (five women and nine men, age range years, mean age 60 years; 54 study objects; one patient examined only once) who were examined with a 1.5 Tesla MR scanner (Siemens, Erlangen, Germany) equipped with a dedicated knee coil. For the coronal turbo SE STIR images, a TR of msec, TE of msec, TI of 130 msec, and flip angle of 90 were used. For the coronal T1w Gd-enhanced fat-suppressed SE images, a TR of 740 msec, TE of 11 msec, and flip angle of 90 were used, with an intravenous contrast media bolus injection of either 12 or 15 ml independently of the patient s body mass. Again, a slice thickness of 4 mm and a acquisition matrix were used for both sequences. Images were also stored in DICOM format, but with a grayscale depth of 11 bits (2048 grayscale values). Image Analysis Quantitative analysis of STIR and T1w Gd-enhanced MR images was performed for both groups 1 and 2 using a recently described computer-assisted method (2). First, a grayscale threshold value (VT) between healthy and edematous bone marrow, based on texture analysis of regions of interest (ROIs) inside healthy bone marrow of three MR slices, was calculated using the computer program MaZda (Lodz Institute of Electronics, Lodz, Poland) (3). Then the examined bone marrow volume of the condyles was segmented manually, slice by slice, using the maximal width of the epiphysis as the volume s maximal height, as seen from the most caudal (for the distal femur) or most cranial (for the proximal tibia) point of the condyles. Areas depicting pathologic changes of the bone marrow other than BME, such as enchondromas, cysts, and circumscribed necrotic lesions, were manually excluded from the bone marrow volume during this step. All pixels inside the thus segmented volume that had a signal intensity higher than the previously calculated VT were treated as BME. Using an EasyVision workstation (Philips, Best, The Netherlands), both absolute volume in cubic centimeters and average absolute signal intensity (ASI) in grayscale values were calculated for the BME. To describe the relative BME volume, the percentage of BME of the examined condyle volume was calculated. To describe the relative BME signal intensity and thus the BME signal contrast, the difference in grayscale values between ASI and VT was used. Only the relative volume and signal contrast of BME were used for further analysis. Statistical Analysis The arithmetic means of relative BME volume and BME signal contrast ( relative BME signal intensity) were calculated independently for the STIR sequence and the T1w Gd-enhanced sequence for each of the two patient groups. In addition, the relative BME volume and BME signal contrast of the STIR and T1w Gd-enhanced sequences were compared by calculating the correlation coefficients (r 2 ) for both patient groups. For each patient in group 2, the quotient of injected contrast media volume in milliliters and body mass in kilograms was calculated. These quotients were then correlated with the differences in relative BME volume and BME signal contrast between the STIR and Gdenhanced MR images. The mean contrast media volume was also determined for patients in group 2. Based on recommendations by Cohen and Holliday (4), the correlation coefficients were interpreted as follows: slight, low, moderate, high, and very high correlation. RESULTS In group 1 (see Table 1), a mean relative BME volume of 15.2% and a mean BME signal contrast (difference ASI to VT) of 70.7 grayscale values were found for the STIR sequence. For the T1w Gd-enhanced fat-suppressed

3 790 Mayerhoefer et al. Table 1 Descriptive Statistics for BME Volume and BME Signal Contrast in Group 1* T1w-Gad STIR r 2 (P 0.01) Mean volume Mean signal contrast *BME volume (% of condyle involvement) and BME signal contrast (difference ASI-to-threshold in grayscale values) in patients of group 1 (16 patients; 0.1 mmol/kg gadopentetate dimeglumine; 1024 grayscale values; early osteonecrosis; N 7; osteoarthritis, N 5; bone bruise, N 3; stress fracture, N 1). T1w-Gad T1-weighted gadolinium-enhanced fat-suppressed MR sequence, ASI average BME signal intensity, r 2 correlation coefficient, VT grayscale threshold value between healthy bone marrow and BME. Table 2 Descriptive Statistics for BME Volume and BME Signal Contrast in Group 2* T1w-Gad STIR r 2 (P 0.01) Mean volume Mean signal contrast *BME volume (% of condyle involvement) and BME signal contrast (difference ASI-to- threshold in grayscale values) in patients of group 2 (14 patients; mmol/kg gadopentetate dimeglumine [mean]; 2048 grayscale values; early osteonecrosis, N 9; osteoarthritis, N 3; stress fracture, N 2). T1w-Gad T1-weighted gadolinium-enhanced fat-suppressed MR sequence, ASI average BME signal intensity, r 2 correlation coefficient, VT grayscale threshold value between healthy bone marrow and BME. sequence, the BME mean relative volume was 16.3% and mean signal contrast was grayscale values (see Fig. 1). Correlation of the two MR sequences was very high for both the relative BME volume and the BME signal contrast. In group 2 (see Table 2), in which a mean contrast media bolus of 0.15 ml ( mmol; range mmol) per kilogram body mass was injected, the mean relative volume of the BME was 9.3% and the mean relative signal intensity was 88.0 grayscale values for the STIR sequence. For the T1w Gd-enhanced fat-suppressed sequence, the mean relative volume was 8.8 % and the mean signal contrast was grayscale values (see Fig. 1). Again, the correlation of the MR sequences was very high for the volume, and at the high end of the high range for the signal contrast. The correlation of the quotients of injected contrast media volume and body mass and the difference in BME volume between the STIR and Gd-enhanced images was moderate (r ; P 0.01; Fig. 2). No significant correlation (r ) was found between the quotients of contrast media volume and body mass and the differences in relative BME signal intensity between the two MR sequences. DISCUSSION It was the aim of this study to compare STIR and T1w Gd-enhanced fat-suppressed MR sequences for depicting BME by computer-assisted quantification, and to determine the influence of injected contrast media volume and variation of MR acquisition parameters on apparent BME volume and signal intensity. BME is an nonspecific lesion that is associated with a variety of disorders, including early-stage osteonecrosis, arthritis, osteoarthritis, stress fractures and stress BME, bone bruise and microfractures, and transient BME syndrome (5). Histology of transient BME syndrome of the hip has revealed diffuse areas of interstitial and intrasinusoidal fluid in bone marrow cavities, as well as fat-cell destruction and regions of fibrovascular regeneration (6). These findings are similar to those reported for early osteonecrosis (7) and BME in osteoarthritic knees (8), and thus a common pathophysiology is discussed. While the term BME seems misleading because the abnormal signal is not (as mentioned above) caused by the presence of fluid alone, it is frequently used in the literature (2,5,7,9 12) and hence is also used in the present study. On MR images, BME typically presents as a diffuse signal abnormality, appearing hypointense on T1w and Figure 1. Bar graph depicting the mean BME volumes (% of condyle involvement) and mean BME signal contrasts (grayscale values between average BME signal intensity and VT) found in groups 1 and 2. Figure 2. Scatter plot depicting the relation of BME volume differences between Gd-enhanced (T1w-Gd) and STIR images and the injected contrast media volumes in group 2.

4 STIR vs. Gadolinium-Enhanced MRI of BME 791 Figure 3. A 57-year-old male from patient group 1, examined at 1.0 Tesla with a Gd bolus injection of 0.1 mmol/kg. Coronal STIR (a) and corresponding T1w Gd-enhanced fat-suppressed (b) MR images depict BME in both the femoral and tibial condyles. Computerassisted quantification of tibial BME (green) revealed a larger volume ( 4.5%) and a higher signal contrast ( 33.2 grayscale values) on Gd-enhanced images (d) than on STIR images (c). hyperintense on T2w MR images. To evaluate the full extent of affected bone marrow on MR images, STIR and T1w Gd-enhanced fat-suppressed MR images are most commonly used, which show increased signal intensity in areas of edematous bone marrow. Despite differences in the MRI techniques used, these sequences have been reported to depict almost identical qualitative and quantitative image patterns, with the BME volume appearing slightly larger on STIR images (1). Apart from tumor BME, in which Gd-enhanced MRI is essential for evaluation of vascularization, it is often a matter of personal choice and financial resources as to whether Gd is used for imaging of BME. Schmid et al (1) were the first to compare BME on STIR and T1w Gd-enhanced fat-suppressed images by assessing BME volume as well as BME signal intensity, which may be equally important in terms of the lesion s clinical course. BME volume can be determined by manual segmentation of the lesion and multiplication by the MR slice thickness, a method that was successfully used to determine the size of osteonecrotic lesions in previous studies (13,14). Signal contrast can be determined by grayscale value analysis of ROIs in edematous and healthy tissue, and calculation of signal-tonoise ratios (SNRs). The main weakness of this method of quantification lies in its observer dependency. Because of the diffuse appearance of BME and its gradual transition to healthy bone marrow, the lesion s true borders are not easily localized by the human eye, which makes manual segmentation difficult. In addition, grayscale value analysis of an ROI inside a diffuse and often multifocal lesion like BME will provide information about the signal intensity of that particular ROI, but not necessarily about the signal intensity of the entire BME. In the present study a recently described, largely observer-independent computer-assisted method for quantifying BME volume and signal contrast (2) was used for comparison of STIR and T1w Gd-enhanced fat-suppressed MR images. The two patient groups were examined at different centers, with each center using a different model of MR scanner operating at a different magnetic field strength. In addition, although each center used the same types of MR images (STIR and Gd-enhanced T1w), they used different routine acquisition parameters and contrast media dosage. Thus, it was possible to evaluate whether variation of major MR parameters significantly influences the difference in apparent BME volume and signal contrast between the two MR sequences. In group 1 patients (Fig. 3), who were examined with a 1.0 Tesla scanner and a contrast media volume of 0.1 mmol per kilogram body mass, Gd-enhanced images showed both higher BME volume ( 7%) and higher BME signal contrast ( 37%) than STIR images. In group 2 patients (Fig. 4), who were examined with a 1.5 Tesla scanner and a mean contrast media volume of mmol per kilogram body mass, BME volume was higher on STIR images ( 5%), while BME signal contrast was again higher on Gd-enhanced images ( 14%).

5 792 Mayerhoefer et al. Figure 4. A 45-year-old male from patient group 2, examined at 1.5 Tesla with an intravenous Gd bolus injection of mmol/kg. Coronal STIR (a) and corresponding T1w Gdenhanced fat-suppressed (b) MR images depict BME in both the femoral and tibial condyles. Computer-assisted quantification of femoral BME (green) revealed a larger volume ( 12.9%) and a higher signal contrast ( 82.0 grayscale values) on STIR images (c) than on Gd-enhanced images (d). These results confirm that BME volume is almost identical on STIR and T1w Gd-enhanced fat-suppressed images, and that Gd-enhanced images show higher BME signal contrast than STIR images. Even though variation of the MR acquisition parameters had no clinically significant effect, the findings differed for the two patient groups. While Gd-enhanced images from patients in group 1 demonstrated a slightly larger BME volume than the STIR images, the opposite was true for patients in group 2. In addition, the difference in BME signal contrast between STIR and Gd-enhanced images was distinctly greater for patients in group 1. These observations might lead to the conclusion that the above results are caused by the difference in injected contrast media volume between group 1 (0.1 mmol/kg) and group 2 (mean mmol/kg). However, Schmid et al (1) also injected contrast media volumes of 0.1 mmol/kg, and found a larger BME volume on STIR images ( 5%) as well as a higher signal contrast on Gd-enhanced images ( 24%). While these authors also found a very high correlation of BME volume on T1w Gd-enhanced fat-suppressed and STIR images, they found only a slight correlation (r ) of BME signal contrast. In the present study, however, a high to very high correlation (r ) of BME signal contrast was observed, a discrepancy that is probably related to the above-mentioned subjective method of quantification used by Schmid et al (1). To further investigate the influence of contrast media volume on the results of the quantification, differences in BME volume and BME signal contrast between STIR and Gd-enhanced images were correlated with quotients of injected contrast media volume and body mass for patients in group 2. Statistical analysis revealed a moderate correlation (r ) between BME volume and contrast media volume. No significant correlation was found between BME signal contrast and injected contrast media volume. While these results suggest that the apparent BME volume is indeed to some degree dependent on the amount of contrast media injected, they do not fully explain the differences observed between patient groups 1 and 2, especially with regard to BME signal contrast. The different T1 relaxation times of tissues and different relaxivities of the contrast agent at both fields (group 1: 1.0 Tesla; group 2: 1.5 Tesla) may be responsible for variations in signal enhancement of the edema and in contrast of the edema to the surrounding tissue. Also, selection of TI (group 1: TI 100 msec; group 2: TI 130 msec) strongly influences the contrast in STIR images. The stronger effect of T2 in STIR images of patient group 2, caused by a longer TE (group 1: TE 10 msec; group 2: TE msec), may also play an important role in this regard because it increases the fluid sensitivity of the sequence. Results from the above-mentioned study by Schmid et al (1) appear to support this theory, because these authors used a TI of 150 msec and a TE of 30 msec for their STIR images, which depicted a slightly larger BME

6 STIR vs. Gadolinium-Enhanced MRI of BME 793 volume than T1w Gd-enhanced images, similar to the findings in our patient group 2. The limitations of this study are mainly related to the relatively small patient population examined. In particular, the influence of injected contrast media volume on the quantitative comparison of MR sequences was investigated in only 27 MR examinations of 14 patients. Patient age was not correlated with the differences in BME volume and signal contrast between the two MR sequences, because no normal distribution of age was observed when the data were graphed. Gender was not evenly distributed in the patient population (only eight females out of 30 patients), and was therefore also not correlated with the results of quantification. In conclusion, the STIR and T1w Gd-enhanced fatsuppressed MR images show a high correlation not only of BME volume, but also of signal contrast, thus contradicting the results of a previous study (1). Variation of major MR acquisition parameters had no significant effect on these findings. The previous suggestion that no substantial difference exists between STIR and Gdenhanced images (1) with regard to the apparent BME volume was confirmed in the present study. Only minor differences were observed between the two study groups. Our data also confirm that Gd-enhanced images provide better signal contrast than STIR images. While a moderate correlation was found between injected contrast media volume and apparent BME volume on Gd-enhanced images, no significant correlation was found between injected contrast media volume and BME signal contrast, indicating that other factors (e.g., the MR acquisition parameters) also influence the apparent imaging pattern of BME on MR images. Because of the higher cost and the small but recognized patient risk of contrast media reaction, Gd-enhanced imaging of BME can be recommended only to confirm a diagnosis, and especially to exclude neoplasm. The STIR sequence is the method of choice for depicting the extent of BME on MR images, despite its lower signal contrast. However, perfusion MRI, which offers better insight into the underlying pathologic processes, may significantly increase the value of contrast media administration for assessing BME. ACKNOWLEDGMENTS MR images were obtained from a clinical study sponsored by Schering AG, Berlin, Germany. REFERENCES 1. Schmid MR, Hodler J, Vienne P, Binkert CA, Zanetti M. Bone marrow abnormalities of foot and ankle: STIR versus T1-weighted contrast-enhanced fat-suppressed spin-echo MR imaging. Radiology 2002:224: Mayerhoefer ME, Breitenseher M, Hofmann S, et al. Computerassisted quantitative analysis of bone marrow edema of the knee: initial experience with a new method. AJR Am J Roentgenol 2004: 182: Mazda, computer program for image texture analysis. Institute of Electronics, Technical University of Lodz, Poland. eletel.p.lodz.pl/cost/progr_mazda_eng.html. 4. Cohen L, Holliday M. Statistics for social scientists. London: Harper and Row; p. 5. Hofmann S, Kramer J, Breitenseher M, Aigner N. Knee pain by bone marrow edema. Arthroskopie 2003:16: Plenk Jr H, Hofmann S, Eschberger J, et al. Histomorphology and bone morphometry of the bone marrow edema syndrome of the hip. Clin Orthop 1997:334: MacDougall L, Conway WF. Controversies in magnetic resonance imaging of the hip. Top Magn Reson Imaging 1996:8: Zanetti M, Bruder E, Romero J, Hodler J. Bone marrow edema pattern in osteoarthritic knees: correlation between MR imaging and histologic findings. Radiology 2000:215: Yoshioka H, Stevens K, Hargreaves BA, et al. Magnetic resonance imaging of articular cartilage of the knee: comparison between fat-suppressed three-dimensional SPGR imaging, fat-suppressed FSE imaging, and fat-suppressed three-dimensional DEFT imaging, and correlation with arthroscopy. J Magn Reson Imaging 2004: 20: Hofmann S, Kramer J, Vakil-Adli A, Aigner N, Breitenseher M. Painful bone marrow edema of the knee: differential diagnosis and therapeutic concepts. Orthop Clin North Am 2004:35: Moosikasuwan JB, Miller TT, Math K, Schultz E. Shifting bone marrow edema of the knee. Skeletal Radiol 2004:33: Felson DT. An update on the pathogenesis and epidemiology of osteoarthritis. Radiol Clin North Am 2004:42: Koo KH, Kim R. Quantifying the extent of osteonecrosis of the femoral head. A new method using MRI. J Bone Joint Surg Br 1995:77: Hernigou P, Lambotte JC. Volumetric analysis of osteonecrosis of the femur. Anatomical correlation using MRI. J Bone Joint Surg Br 2001:83: