Detection of malignant bone marrow involvement with dynamic contrast-enhanced magnetic resonance imaging

L. A. Moulopoulos1,+, T. G. Maris3, N. Papanikolaou3, G. Panagi4, L. Vlahos1 and M. A. Dimopoulos2

Departments of 1 Radiology and 2 Clinical Therapeutics, Medical School, University of Athens, Athens; 3 Department of Medical Physics, University Hospital of Heraklion, Heraklion, Crete; 4 Department of Radiology, General Hospital of Chios, Chios, Greece

Received 21 March 2002; revised 5 June 2002; accepted 17 July 2002


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Background:

The purpose of this study was to evaluate the role of dynamic contrast-enhanced magnetic resonance imaging (dMRI) in detecting bone marrow involvement in cancer patients.

Patients and methods:

We studied 50 consecutive patients with histologically confirmed malignant dissemination to the bone marrow, using dMRI of the lumbosacral spine. Time–signal intensity curves were generated from regions of interest (ROIs) obtained from areas of obvious bone marrow disease (group B). In 16 patients from group B with focal disease, ROIs were also placed on areas with apparently normal bone marrow on static magnetic resonance images (group C). Twenty-two patients with no history of malignancy were used as a control group (group A). Wash-in (WIN) and wash-out (WOUT) rates, time to peak (TTPK), time to maximum slope (TMSP) values and WIN/TMSP ratios were calculated for each patient. Mean values for the three groups were compared statistically. Six patients from group B had follow-up dMRI after chemotherapy: four patients achieved a clinical partial response and two had resistant disease.

Results:

A significant difference was found between groups A and B for all values. Between groups A and C, in spite of the similar static MRI appearance, all values were significantly different. Between groups B and C, a significant difference was found for WIN, WOUT rates and WIN/TMSP ratio. Follow-up dMRI data analysis correlated well with clinical staging.

Conclusions:

dMRI can distinguish normal from malignant bone marrow. It may identify malignant bone marrow infiltration in patients with negative static MRI and serve as both a diagnostic and prognostic tool for patients with bone marrow malignancies.

Key words: bone marrow, dynamic magnetic resonance imaging, magnetic resonance imaging


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Magnetic resonance imaging (MRI) has contributed greatly to the detection and assessment of bone marrow malignancies [1, 2]. A conventional MRI study of the bone marrow includes relatively T1- and T2-weighted images, and is usually sufficient to establish a diagnosis of malignancy in the presence of abundant fatty marrow. Because T1 and T2 values of some tumors approximate those of hematopoietic bone marrow, it may be difficult to detect tumor when red marrow predominates in the skeleton.

The absence of a single predictable pattern of red to yellow marrow conversion in the spine complicates the search for malignant bone marrow lesions further [3]. The institution of chemotherapy and, in particular, the concomitant use of growth factors leads to an increase in the volume of red marrow. In addition to red marrow hyperplasia, other disorders of the bone marrow that can complicate the course of disease in a cancer patient (fibrosis, infarction, edema related to recent compression fractures, infection) may also simulate tumor.

Dynamic contrast-enhanced magnetic resonance imaging (dMRI) studies the kinetics of the distribution of paramagnetic contrast in the microvessels and in the interstitial space of the tissues being studied. dMRI has been applied to the study of musculoskeletal tumors, with encouraging results regarding the diagnosis of malignant from benign and viable from non-viable tumor, and the prediction and assessment of response to chemotherapy [4–9]. We undertook this study to investigate the potential of dMRI to separate malignant bone marrow disease from uninvolved bone marrow, and its ability to detect underlying disease in patients with known bone marrow malignancies and normal conventional MRI studies.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Patients
Fifty consecutive patients with pathologically confirmed malignant disease of the bone marrow were enrolled in this study (group B; Table 1). Patients’ ages ranged from 24 to 80 years (mean 62 years), and the group comprised 22 men and 28 women.


View this table:
[in this window]
[in a new window]
 
Table 1. Primary malignancies of patients with abnormal bone marrow (group B)
 
Twenty-two consecutive patients (14 men, 8 women), who were referred for evaluation of degenerative disc disease and who had no known history of malignancy, served as a control group (group A). Patients’ ages ranged from 29 to 72 years (mean 60 years). All patients underwent dMRI studies of the bone marrow. MRI studies for group B were performed prior to initiation of treatment.

Six patients (one patient each with breast cancer, lung cancer, cervical cancer and myelodysplastic syndrome, and two with multiple myeloma) had follow-up dMRI studies within a month from completion of therapy.

Oral consent was obtained from all patients.

MR imaging
Magnetic resonance images were obtained with a 1.5 T unit (Philips Medical Systems, Eindhoven, The Netherlands). Sequences for static MRI included sagital T1-weighted SE [repetition time (TR) = 500–600 ms, echo time (TE) = 10–20 ms] and turbo short-time inversion recovery [STIR; TR = 1620 ms, inversion time (TI) = 180 ms, TE = 70 ms, turbo spin echo factor = 12) images of the lumbosacral spine. Imaging parameters were: 4 mm section thickness, 1 mm interslice gap, 204 x 256 imaging matrix, 2–3 signals averaged. dMRI was performed with a T1-weighted gradient-echo sequence (TR = 11 ms, TE = 4.2 ms, flip angle = 30°). The dynamic study was limited to a single spinal part to avoid inherent variations in perfusion between different spinal locations. The lumbosacral spine was selected for dMRI over other skeletal parts to match the patients’ studies with those of the control group.

Five sections through the spine, carefully selected from the static study to encompass areas of abnormal marrow, were obtained every 18 s for a total of a maximum of 3 min. A bolus of gadopentetate dimeglumine 0.1 mmol/kg body weight (Magnevist; Schering, Berlin, Germany) was injected manually immediately after the end of the first dynamic acquisition.

After completion of the MRI study, all contrast-enhanced dynamic images were subtracted from the first set of unenhanced images by using the subtraction function of the magnetic resonance (MR) unit. For each normal control of group A, a region of interest (ROI) was positioned on one of the lumbar vertebral bodies, avoiding the areas affected by motion artifacts. For each patient of group B, subtraction images were reviewed and an ROI was placed on a focus of maximum abnormal enhancement. Foci of abnormal enhancement that could be related to vertebral endplate perfusion changes that accompany degenerative disc disease were excluded from the study. All ROIs measured between 70 and 80 pixels, and were carefully selected to avoid the basivertebral vessels that course at the midline of each vertebra. ROIs were selected on a single image and were then plotted automatically on all images of the same dynamic series. In 16 patients from group B in whom the neoplastic process did not appear to involve the entire bone marrow on the static MR images, a second ROI was positioned at a site of apparently normal marrow (group C). For the six patients from group B who had follow-up dMRI studies, ROIs were positioned on the same vertebra as during the pre-treatment study.

Data analysis
The dynamic signal enhancement (DSE) of selected ROIs was plotted as a function of time (t) and signal intensity–time curves were generated (Figures 1, 2 and 3). Signal intensity–time data obtained from each ROI were fitted by means of a Marquardt linear regression analysis method [10] according to the formula:



View larger version (33K):
[in this window]
[in a new window]
 
Figure 1. Mean DSE and mean first derivative d(DSE)/dt curves. Calculated fit parameters were: a = 70.70, b = 74.37, c = 8.510, d = 65.03; r = 0.9963; Fstat = 228. WIN, WOUT, TTPK and TMSP parameters were calculated from the d(DSE)/dt curve and their numerical values are: WIN = 2.19 s1, WOUT = –0.120 s1, TTPK = 74.37 s and TMSP = 38.46 s. WIN/TMSP = 0.057 s2.

 


View larger version (34K):
[in this window]
[in a new window]
 
Figure 2. Mean DSE and mean first derivative d(DSE)/dt curves. Calculated fit parameters were: a = 365.83, b = 53.86, c = 4.793, d = 96.63; r = 0.9998; Fstat = 4937. WIN, WOUT, TTPK and TMSP parameters can be easily calculated from the d(DSE)/dt curve and are presented on the graph. Their numerical values are: WIN = 19.86 s1, WOUT = –0.753 s1, TTPK = 53.86 s and TMSP = 31.80 s. WIN/TMSP = 0.624 s2.

 


View larger version (28K):
[in this window]
[in a new window]
 
Figure 3. Mean DSE and mean first derivative d(DSE)/dt curves. Calculated fit parameters were: a = 115.78, b = 55.78, c = 6.022, d = 49.31; r = 0.9961; Fstat = 211. WIN, WOUT, TTPK and TMSP parameters were calculated from the d(DSE)/dt curve and their numerical values are: WIN = 5.13 s1, WOUT = –0.360 s1, TTPK = 55.91 s and TMSP = 31.97 s. WIN/TMSP = 0.160 s2.

 


(1)

Parameters (a, b, c, d) were calculated from the fit of DSE to t.

For all fits, r was >=0.9. This value was used as a threshold for the estimation of goodness of each fit. Wash-in (WIN) and wash-out (WOUT) rates were calculated for each patient from the maximum and minimum slopes of signal intensity–time curves, respectively. These slopes, as well as time to peak (TTPK) and time to maximum slope (TMSP) values were obtained from the first derivative function f(t) = d(DSE)/dt of equation (1). The ratio WIN/TMSP was also calculated.

Patients and normal controls were divided into three sample groups for statistical evaluation of the results: (i) group A (normal controls); (ii) group B (patients with abnormal bone marrow); and (iii) group C (patients from group B with areas of normal-appearing bone marrow on static MR images). Assuming the possibility of difference in standard deviations amongst compared populations, an unpaired t-test with Welch correction was used to check statistical significance amongst all groups (sample means) using WIN, WOUT, TTPK, TMSP values and the ratio of WIN/TMSP. A P value of 0.05 was considered a statistically significant threshold.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Mean and median values of calculated parameters WIN, WOUT, TTPK, TMSP and WIN/TMSP for the three study groups (A, B and C) are presented in Table 2. Statistical results (P values) are shown in Table 3. Graphical presentations of mean DSE and mean first derivative d(DSE)/dt curves were produced from the relevant patient samples for all three study groups (Figures 13).


View this table:
[in this window]
[in a new window]
 
Table 2. Measured mean and median WIN, WOUT, TTPK, TMSP values and WIN/TMSP ratios in three subject sample groups
 

View this table:
[in this window]
[in a new window]
 
Table 3. Statistical results (P values) using unpaired t-test with Welch correction for differences between mean WIN, WOUT, TTPK and TMSP values, and WIN/TMSP ratios amongst sample groups
 
Normal compared with abnormal bone marrow (groups A and B). Mean WIN and WOUT rates and WIN/TMSP ratios were significantly lower for group A compared with those of group B. Mean TTPK and TMSP values were significantly higher for normal subjects (group A) than for patients with abnormal bone marrow (group B).

Abnormal compared with normal-appearing bone marrow (groups B and C). Mean WIN and WOUT rates and WIN/TMSP ratios were significantly higher for the group with abnormal bone marrow (group B) compared with those of the group with apparently normal bone marrow on static MR images (group C).

Mean TTPK and TMSP values for groups B and C did not differ significantly.

Normal compared with normal-appearing bone marrow (groups A and C). Mean WIN rates and WIN/TMSP ratios were significantly shorter for group A compared with those of group C.

Mean TTPK and TMSP values were significantly higher for group A when compared with group B.

Follow-up MRI
Four of six patients with malignant bone marrow disease who had follow-up MRI achieved a clinical partial response to treatment, while in two patients the disease progressed.

dMRI diagnosis was consistent with the clinical evaluation in all patients (Tables 4 and 5; Figure 4).


View this table:
[in this window]
[in a new window]
 
Table 4. Follow-up MRI: pre-treatment dMRI data in six patients
 

View this table:
[in this window]
[in a new window]
 
Table 5. Follow up MRI: post-treatment dMRI data in six patients
 


View larger version (139K):
[in this window]
[in a new window]
 
Figure 4. A 76-year-old woman with breast cancer. Pre-treatment T1-weighted spin echo (TR = 600 ms, TE = 20 ms) (A) and dynamic gradient echo (TR = 12 ms, TE = 4.2 ms, FA = 30° at 49 s) (B) sagital magnetic resonance images show diffuse low signal intensity (A) and marked enhancement (B) of abnormal bone marrow. Post-treatment T1-weighted (C) and dynamic gradient echo (D) sagital magnetic resonance images obtained after completion of chemotherapy show persistent abnormal bone marrow signal in (C), but no enhancement in (D).

 
Static post-treatment MR images were in accordance with the clinical diagnosis in only two of six patients; in four patients no change in the appearance of the bone marrow lesions could be detected on the follow-up study.


    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Although MRI is the most sensitive imaging method for detecting bone marrow abnormalities, it is still difficult to identify these lesions when the bone marrow is richly cellular, as in infants and children, and in adults with hyperplasia of the red marrow. Red marrow does enhance after the administration of paramagnetic contrast, but to such a degree that, in adults, there is faint, if any, visually detectable change in its signal intensity. In dMRI studies, red marrow has a fairly characteristic contrast profile with early enhancement, which peaks within 1 min from contrast injection.

Dynamic contrast-enhanced MRI has been applied extensively to the study of solid musculoskeletal tumors, but there are very few publications on dMRI of bone marrow. Bollow et al. studied 30 patients with B-cell-type chronic lymphocytic leukemia (B-CLL) and 45 patients without known malignancy with dMRI of the bone marrow, and found significant differences between patients with Binet stages B and C of B-CLL and normal controls [11]. Hawighorst et al. reported differences in pharmacokinetics between 10 patients with multiple myeloma and seven healthy controls who were studied with dMRI of the bone marrow [12]. Both studies show that there are significant differences between the contrast profiles of normal bone marrow and those of hematopoietic malignancies.

Our study showed significant differences between signal intensity–time curves obtained from normal controls (group A) and those obtained from patients with malignant disease of the bone marrow (group B). When compared with red marrow, enhancement of abnormal bone marrow occurred earlier, was steeper and did not last as long. All four parameters, calculated from the fit of the dynamic signal enhancement to time, differed significantly between the two groups (Table 3). The higher WIN and WOUT, and the lower TTPK and TMSP values of malignant bone marrow compared with those of normal bone marrow in our study reflect the faster throughput of contrast material, which can be explained by the presence and number of abnormal vessels with increased size and permeability [13].

Before a malignant bone marrow lesion becomes apparent on a conventional MRI study, it must replace enough normal bone marrow cells to cause a local alteration of T1 and T2 values; in cases of early dissemination to the bone marrow, a conventional MRI study may indeed be normal. Because an abnormal dMRI signal intensity–time curve depends on contrast medium kinetics, it may detect tumor neovascularity before a lesion becomes apparent on conventional MRI. In our study, signal intensity–time curves obtained from areas of bone marrow that did not appear to be involved on pre-contrast MR images of patients with known malignant dissemination to the bone marrow (apparently normal bone marrow; group C) were clearly different from those obtained from healthy controls (group A); quantitative analysis confirmed this observation by showing significant differences in all four parameters calculated from the fit of the dynamic signal enhancement to time, and in the WIN/TMSP ratio, which reflects the speed of change of WIN values (Table 3).

When apparently normal bone marrow signal intensity–time curves (group C) were compared with those of abnormal bone marrow (malignant bone marrow; group B), a significant difference was found only between WIN and WOUT values and WIN/TMSP ratios. However, TTPK and TMSP values of apparently normal bone marrow approximated those of malignant bone marrow. Dynamic MRI profiles of apparently normal bone marrow (group C) bared a closer resemblance to those of abnormal bone marrow (group B) than to those obtained from healthy controls. Whether the difference in contrast kinetics is due to the presence of neovascularity related to tumor angiogenesis or whether it could be related to increased perfusion of uninvolved bone marrow that is located adjacent to a site of tumoral infiltration remains to be seen. Hawighorst and colleagues reported significantly higher contrast uptake in patients with monoclonal gammopathy of unknown significance (MGUS) compared with normal controls, even though a distinction between the two groups based on visual interpretation of the conventional MRI study was not possible [12]. It seems, therefore, that dMRI of the bone marrow may show changes during the early phase of tumor angiogenesis and, thus, diagnose bone marrow involvement before it becomes apparent on conventional MR images. Correlation of dMRI data with histopathological parameters related to tumor angiogenesis may confirm the above hypothesis.

If dMRI can distinguish red marrow from abnormal marrow, one of its potential applications is the detection of residual disease in patients who are being treated for malignant disease of the bone marrow. In all six patients who had follow-up dMRIs in our study, dMRI data were in accordance with the clinical evaluation of response (Tables 4 and 5). Conversely, conventional post-treatment MR images failed to show any change from the baseline study in four out of six patients. In one patient with myelodysplastic syndrome, dMRI correctly diagnosed response to treatment in spite of the absence of abnormal findings in both pre- and post-treatment conventional MRI studies. It is possible that dMRI may detect the presence or absence of changes indicative of response to treatment before these become evident on the static MR images, providing, therefore, the attending physician with an early assessment of treatment.

Clearly, our data applies only to the differential diagnosis of malignant from uninvolved red marrow, and further studies comparing signal intensity–time curves of malignant bone marrow infiltration with those of benign bone marrow abnormalities or endplate abnormalities associated with degenerative disc disease may be relevant. The latter were carefully avoided in our study because increased bone marrow perfusion related to their presence could alter the contrast uptake profile of uninvolved, normal bone marrow. However, most of the benign disorders that may complicate the course of treatment in a cancer patient have morphological characteristics that should enable their diagnosis on pre-contrast MR images.

Dynamic MRI is easily performed and the software is available in most currently used MR units. We have shown that dMRI can distinguish normal from abnormal bone marrow. Mapping of normal bone marrow contrast profiles in different age groups, and dMRI studies of patients with the same bone marrow malignancy and correlation of dMRI data with parameters of the histopathologic specimen may not only establish the role of dMRI in the work-up of bone marrow malignancies, but may also provide information on the development and evolution of bone marrow malignancies from the early phase of angiogenesis.


    Footnotes
 
+ Correspondence to: Dr Lia A. Moulopoulos, 4 Ivis Street, Ekali, Athens 14565, Greece. Tel: +30210-6228771; Fax: +30210-8131383; E-mail: lia1312{at}otenet.gr Back


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
1. Vogler JB, Murphy WA. Bone marrow imaging. Radiology 1998; 168: 679–693.[ISI][Medline]

2. Moulopoulos LA, Dimopoulos MA. Magnetic resonance imaging of the bone marrow in hematologic malignancies. Blood 1997; 90: 2127–2147.[Free Full Text]

3. Ricci C, Cova M, Kang YS et al. Normal age-related patterns of cellular and fatty bone marrow distribution in the axial skeleton: MR imaging study. Radiology 1990; 177: 83–88.[Abstract]

4. Erlemann R, Reiser MF, Peters PE et al. Musculoskeletal neoplasms: static and dynamic Gd DTPA-enhanced MR imaging. Radiology 1989; 171: 767–773.[Abstract]

5. Erlemann R, Sciuk J, Bosse A et al. Response of osteosarcoma and Ewing sarcoma to preoperative chemotherapy: assessment with dynamic and static MR imaging and skeletal scintigraphy. Radiology 1990; 175: 791–796.[Abstract]

6. Fletcher BD, Hanna SL, Fiarclough DL, Gronemeyer SA. Pediatric musculoskeletal tumors: use of dynamic contrast-enhanced MRI imaging to monitor response to chemotherapy. Radiology 1992; 184: 243–248.[Abstract]

7. Reddick WE, Bhargava R, Taylor JS et al. Dynamic contrast-enhanced MRI evaluation of osteosarcoma response to neoadjuvant chemotherapy. J Magn Reson Imaging 1995; 5: 689–694.[Medline]

8. Verstraete KL, de Deene Y, Roels H et al. Benign and malignant musculoskeletal lesions: dynamic contrast-enhanced MR imaging-parametric ‘first-pass’ images depict tissue vascularization and perfusion. Radiology 1994; 192: 835–843.[Abstract]

9. Van der Woude HJ, Verstraete KL, Hogendoorn PCW et al. Musculoskeletal tumors: does fast dynamic contrast-enhanced subtraction MR imaging contribute to characterization? Radiology 1998; 208: 821–828.[Abstract]

10. Marquardt DW. An algorithm for least squares estimation of non-linear parameters. J Soc Industr Appl Math 1963; 2: 431–441.

11. Bollow M, Knauf W, Korfel A et al. Initial experience with dynamic MR imaging in evaluation of normal bone marrow versus malignant bone marrow infiltrations in humans. J Magn Reson Imaging 1996; 7: 241–250.

12. Hawighorst H, Libicher M, Knopp MV et al. Evaluation of angiogenesis and perfusion of bone marrow lesions: role of semiquantitative and quantitative dynamic MRI. J Magn Reson Imaging 1999; 10: 286–294.[CrossRef][ISI][Medline]

13. Buada LD, Mukarami J, Murayama S et al. Breast lesions: correlation of contrast medium enhancement patterns on MR images with histopathologic findings and tumor angiogenesis. Radiology 1996; 200: 639–649.[Abstract]