MRI and CT Evaluation of Primary Bone and Soft- Tissue Tumors

749 Alex M. Aisen1 William Martel1 Ethan M. Braunstein1 Kim I. McMillin1 William A. Phillips2 Thomas F. KIing2 Received June 10, 1985; accepted after revision December 23, 1985. Presented at the annu meeting ofthe American Roentgen Ray Society, Boston, April 1985. 1 Department of Radiology, lhversity of Michigan Hospitals, Ann Arbor, Ml 48109. Address reprint requests to A. M. Aisen (Box 13). 2p of Surgery, University of Michigan Hospitals, Ann Arbor, MI 48109. AJR 146:749-756, April 1986 0361-803x/86/1 464-0749 American Roentgen Ray Society MRI and CT Evaluation of Primary Bone and Soft- Tissue Tumors Twenty-six patients with primary tumors of bone or somatic soft tissues underwent both magnetic resonance imaging (MRI) and computed tomography (CT); 15 of the patients had radionuclide bone scans as well. Only in a minority of cases did these tomographic methods provide information needed for diagnosis thatcould not be derived from the plain radiographs alone; however, for assessing the extent of the disease, both CT and MRI proved very valuable, particularly MRI. Specifically, MRI was superior to CT in delineating the extent of the neoplasms and their relation to surrounding structures in 21 of the patients, equal in four, and interior in only one. Furthermore, in the 13 patients with tumors of long bone, MRI was judged superior to CT in visualizing marrow abnormalfty in 12 cases, and equal in only one case. Radionudhde scans demonstrated the lesions in 14 of the 15 cases; its primary utilfty was in excluding additional lesions. It is concluded that for these patients, MRI was the imaging method of choice in assessing the extent of bone and soft-tissue tumors. Magnetic resonance imaging (MRI) is proving to be a valuable technique in the evaluation of patients with primary tumors of the bones and somatic soft tissues [1-4], particularly because of the high contrast between tumors and normal tissues generally found with MRI. In this prospective study, we describe the relative efficacy of MRI, performed with a superconducting magnet, compared to CT in diagnosing and evaluating the visibility and extent of disease in 26 patients with tumors of the bone and somatic soft tissues. Subjects and Methods Twenty-six patients with new or recurrent primary tumors of bone or somatic soft tissues were included in this prospective study. All patients who presented at the University of Michigan Hospital with such lesions between January 1984 and August 1985, and who had both x-ray CT and MRI examinations and histologically proven diagnoses from either surgical or biopsy specimens, are included in this report. There were 15 female and 1 1 male patients, with a mean age of 26 years (range, 2-80). Sixteen patients had malignant lesions, and 10 had benign tumors. The neoplasms arose in osseus tissue in 19 and in soft tissue in seven individuals. Pathologic verification of the extent of disease in the bone marrow following resection was available in four cases. In addition to MRI and CT, plain films were available for all patients, and radionuclide bone scans for 15. The MRI and CT studies were reviewed by at least two of us, one primarily a bone radiologist, the other specializing in MRI. The MRI and CT examinations were first correlated with the plain films to determine whether the tomographic studies added any additional information needed for diagnosis. The MRI studies were then compared with the CT examinations to judge relative efficacy in evaluating the extent of disease. Ratings of from 0 (poorest) to 4 were assigned to MRI and CT in four categories, on the basis of a consensus of the two observers. Not every category was applicable to each case; the number of cases for which each was applicable is shown in brackets. The categories were (1) clarity of depiction of the boundary between the lesion and surrounding normal tissues [26 cases]; (2) contrast difference between the tumor and surrounding soft tissues [23 cases]; (3) ability to assess the relation of the abnormal tissue

750 AISEN ET AL. AJR:146, April 1986 to surrounding nerves and/or blood vessels [20 cases]; and (4) ability to assess extent of abnormality in the bone marrow [1 3 cases]. CT and MRI were then compared case by case on the basis of the grades in each individual category, and again using the average grade of all applicable categories for each case. The results were then totaled for the 26 patients, to produce a relative comparison of CT and MRI for each category, as well as for the average of all four categories. The radionuclide studies were reviewed to determine if they provided additional information. Imaging Techniques MRI was performed on a Diasonics MT/S system, based on a 0.35-T superconducting magnet, using multislice dual spin-echo (SE) pulse sequences, with echo delay times (TE) of 28 and 56 msec. Slices were 8 mm thick, with a 2-mm gap between adjacent images; the pixel size was about 1.7 x 1.7 mm. Whole-body-sized imaging coils were used for all patients, except three whose tumors involved the head and neck; a standard head coil was used in these cases. Pulse repetition (TR) intervals were in the range 500-2000 msec, with scanning performed at a minimum of two TR intervals, to permit calculation of Ti and T2 relaxation times. All patients but one were imaged using transverse tomographic slices; coronal and/or sagittal images were available in 18 of the 26 patients. CT was performed on GE 9800 or 8800 scanners; intravenous contrast material was used in nine patients. One of the CT studies was performed at another institution. Radionuclide studies were performed using routine technique, after injection of 15 mci (555 MBq) (less in children) of technetium-99m methylene diphosphonate. Seven of the patients underwent repeat MRI examination after chemo- and/or radiotherapy. Ti and T2 relaxation times were calculated for tumor, muscle, and fat using software furnished with the scanner. Analysis was based on sets of images at the same level, acquired using two TR and two TE intervals. Results Diagnostic Utility B The MRI studies provided information useful for diagnosis, but not present on plain radiographs, in seven patients; this was true of CT in five patients, who were also among these seven. In four of these five patients this stemmed from the ability to identify fatty tissue, shown equally well by MRI and CT; in the fifth, an aneurysmal bone cyst, a fluid-fluid level was found with both methods. In the remaining two patients, both with aggressive fibromatosis, the appearance of the lesions on MRI was distinctive; the lesions had very low signal intensities on both short and long TR and TE intervals (fig. 1). This is consistent with a low mobile proton density, as one would expect in a dense fibrous lesion, and in this regard the scans provided useful diagnostic information. There were two cases (aneurysmal bone cyst and juxtacortical chondroma) where the demonstration of a thin ossific rim on CT, suggesting confinement of lesion to bone, provided information absent on both plain films and MRI. However, it was not believed that this information changed the differential diagnosis that would have been derived from the plain films alone. Accuracy Inasmuch as few patients underwent surgical resection of the neoplasm, we did not have an adequate basis for judging the accuracy with which these imaging methods depicted the true extent of tumor in the soft tissues and bone marrow. Further, we were usually unable to determine with certainty

AJR:146, April 1986 MRI/CT OF BONE & SOFT-TISSUE TUMORS 751 Fig. 2-14-year-old boy with osteosarcomaofdistal rightfemur. A, Plain film. B, Coronal SE 500/28 image through losion. Note area of presumed hemorrhage secondary to biopsy (arrow) with high signal intensity. Lesion does not cross physis and has sharp boundary proximally. C, SE 2000/28 in similar section. 0, SE 500/28 in similar section 2 months later, after chemotherapy, revealing little change in tumor size. E, SE 2000/28 same time as D. F, Pathologic specimen. Surgery was performed 2 days after images shown in 0 and E were acquired. Note sharp cutoff (arrow); no tumor was found proximally; epiphysis was also tumor free.

752 AISEN ET AL. AJR:146, April 1986 whether the abnormal signals present on MRI represented tumor or edema. However, in the four patients in this series (three with osteosarcoma and one with giant cell tumor) who underwent resection within 3 days of MRI, we were able to verify that MRI had accurately portrayed lesion extent (figs. 2 and 3). The marrow extent in these patients and the soft tissue extent in the three with osteosarcoma was seen less clearly on CT than on MRI. Figure 4 demonstrates a patient in whom an area of abnormal signal seen in the soft tissues adjacent to the bone proved to contain tumor; however, we have no basis for concluding that this will always be the case, and suspect that in some cases areas of abnormal signal will simply represent edema. Demonstration of Disease Extent The results of the comparison between MRI and CT are summarized in table 1. MRI was judged superior to CT in the majority of cases in all individual categories, as well as in the overall rating. The major reason was the greater contrast between neoplasm and adjacent normal tissue, generally fat and muscle, characteristic of MRI scans compared to CT (fig. 5). In several patients we were able to visualize thin rims of sclerotic bone on CT not visible on MRI. However, in these cases this information did not add significant diagnostic information to that already available from the plain radiographs. In the two patients with retroperitoneal neoplasms, MRI was hindered because of the lack of a bowel contrast agent; motion also produced image degradation. The CT scans in nine patients were contrast enhanced. In four of these the enhancement significantly improved the utility of the CT examinations, generally by rendering blood vessels more visible; contrast of the lesion with muscle also improved in one patient with aggressive fibromatosis. However, in only two of the cases did the use of contrast material change the overall ratings. Although a properly timed bolus Fig. 3.-i 6-year-old girl with osteosarcoma of left distal femur. A, Plain film. B, Coronal SE 500/28 image. Lesion involves epiphysis, is sharply demarcated, and has more intense signal than muscle. C, Coronal SE 500/28 image 3 months later, after chemotherapy. Lesion is smaller and has become isotense with muscle. Note biopsy sites (arrows). of intravenous contrast material can be helpful with CT, it may also add ambiguity in the identification of calcium and does carry a slight risk of contrast reaction. It was generally helpful to obtain both short (usually 500 msec) and long (1 500 or 2000 msec) TA interval images on MRI. The former provided the best technique for differentiation of tumor and fat (including marrow fat) and the latter was best for differentiation of tumor and muscle. Coronal and sagittal views were also generally very helpful in assessing the extent of lesions. It was generally easier to judge the relation of tumors to vital structures, such as blood vessels or spinal cord, with MRI than with CT (fig. 6). Marrow Extent Of the 1 3 patients with tumor of a long bone, MRI was superior to CT in 12 cases and equal in only one in demonstrating the extent of abnormal bone marrow. Such assessment was best done on short-ta scans. Imaging in the coronal or sagittal planes was also valuable, and provided a significant advantage over CT. Figures 2 and 3 show patients with osteosarcoma in whom MRI clearly demonstrated the extent of abnormality in the bone marrow. In both cases the diaphyseal boundaries of the marrow abnormality on the posttherapy MRI scans corresponded well to the tumor margins found in the sectioned surgical specimens. In the patient shown in figure 2, the lesion does not cross the epiphyseal plate (confirmed histologically); in the other patient extension into the epiphysis is evident. The absence of beam-hardening artifacts in MRI was a decided advantage over CT (fig. 7). Response to Therapy Of the seven patients scanned before and after chemo- or radiotherapy, significant reductions in lesion size were noted in three (two Ewing tumors and one osteosarcoma; figs. 3

AJR:146, April 1986 MRI/CT OF BONE & SOFT-TISSUE TUMORS 753 Fig. 4.-i 3-year-old girl with osteosarcoma of proximal right tibia. Coronal SE 2000/28 image. Increased signal in muscle (arrow), which proved to contain viable neoplasm histologically. Fig. 6-45-year-old man with plasmacytoma of T6 vertebra, involving body and pedicle. Coronal SE 500/28 image reveals impingement on spinal cord. On this short-tr, Ti -weighted image, spinal fluid in dural sac is not seen. and 5); in two of these there were also posttherapeutic changes in relative signal intensity of the neoplasm compared to adjacent normal tissues. In three of the remaining four cases (two osteosarcornas and one chondrosarcoma) the MRI appearance did not change substantially (fig. 2). In one case, a patient with osteosarcoma, a sharp boundary developed between the intrarnedullary part of the lesion and the adjacent normal marrow after therapy, while the rest of the tumor did not change in appearance substantially. The tumors in four of the seven cases were evaluated ---- _ -- nply fig. 4). C, Coronal image clearly demonstrates extent of disease. 0, Transverse SE 2000/284 months later, after radiation and chemotherapy, shows regression. histologically both before and after therapy. In all cases the response to therapy was gauged histopathologically on the basis of tumor necrosis. One patient who showed a reduction in tumor size as well as a change in intensity characteristics of the neoplasm on MRI showed good response to chemotherapy pathologically (fig. 3). A second patient, whose lesion did not change on MRI, showed a poorer response to chemotherapy histopathologically (fig. 2). The third patient, who showed only a slight change in the size and intensity of the lesion on MRI, exhibited minimal response to therapy. In the

754 AISEN ET AL. AJR:146, April 1986 TABLE 1: Comparison of Lesion Depiction on MRI and CT MR <CT MRI=cT MRI>cT Depiction of tumor boundary 5 7 i 3 Contrast between tumor and surrounding soft tissues 0 6 17 Demonstration of relation of tumor to adjacent nerves and vessels 2 7 11 Visualization of bone marrow abnormality 0 i 12 Overall rating i 4 21 final patient, in whom only the intramedullary part of the lesion changed, there was also only limited response to therapy. Calculated Relaxation Times Ti and T2 relaxation times for normal muscle and fat and for the tumors are shown in table 2. The range of relaxation times for the tumors was great; the values were not helpful in characterizing tumors by tissue type or in distinguishing benign from malignant lesions. Radionuclide Studies Radionuclide bone scans were performed in 1 5 of the patients, and were positive in 1 4. Aadionuclide examination demonstrated the bony extent of the 1 4 lesions clearly, but spatial resolution was poor. None of the bone scans added information about tumor extent to that already available from CT and MRI. A major advantage of radionuclide imaging was in excluding additional remote skeletal lesions. Discussion The value of MAI in evaluating tumors of the musculoskeletal system has been previously noted [1-4]. MRI offers great contrast between normal and abnormal tissues. The absence of signals from cortical bone and calcium is a hindrance in diagnosis, but a virtue in permitting assessment of tumor extent, unimpaired by beam hardening and other bone-related artifacts found in x-ray imaging. In two recent series comparing the efficacies of MRI and CT, MRI was found to be of equal or greater value than CT in assessing the extent of disease in the majority of cases [3, 4]. Our findings show MRI to be the superior method in a much greater percentage of cases than in these reports. This may be related to our having used an MRI system based on a superconducting magnet, whereas the previous two groups used lower-field-strength resistive systems. Our results also suggest that there is an important role for MRI in evaluating response to nonsurgical therapy. Taking all factors into account, we found MRI superior to CT in delineating disease in the large majority of our cases. Several operational factors were important in realizing this advantage, including the use of coronal or sagittal as well as transverse imaging planes. Also, the choice of proper pulse sequences is important. We found SE sequences excellent in the evaluation of these neoplasms. It was often necessary to use at least two sequences: a short-ta sequence for maximal contrast between lesion and fat, including bone marrow fat, and a long-ta sequence for optimal contrast between the lesion and adjacent muscle. Coronal and sagittal images were of great help in demonstrating tumor extent. For tumors involving the long bones, scanning is generally best performed with the affected limb placed parallel to the imaging plane. Motion artifacts and a lack of a suitable bowel contrast agent are limitations of MAI in the abdomen; both problems may yield to technologic advances. Although we found MRI to be subjectively superior in delineating the extent of tissue abnormality, the accuracy of lesion depiction is obviously a significant issue. We had only four cases in which accuracy was pathologically verified, and in these cases found MAI to be correct. It is inappropriate to draw strong conclusions from so few cases, but we believe MAI will indeed prove to be as accurate, if not more so, than CT in the delineation of disease extent. Furthermore, MAI s clear delineation of abnormal tissue and its relation to adjacent vascular and nervous structures suggests that it is more useful than other currently available imaging methods in preoperative staging, planning the surgical approach, and determining the potential for limb salvage. A related issue concerns verification of the nature of abnormal tissue identified on MRI. We recognize, for example, that it can be difficult to distinguish reactive edema from neoplasm using current imaging techniques, including MRI. This question, too, will need to be addressed in further studies that provide radiologic-pathologic correlation. We have also called attention to the potential value of MRI in gauging response to therapy. Again, it is not possible to draw firm conclusions on the basis of our limited experience, although our findings are encouraging. In addition to obtaining radiologic-pathologic correlation in many more cases, issues such as the number, timing, and location of biopsy sites must be addressed. Neither MRI nor CT was especially useful in providing diagnostic information not demonstrable on plain films. The two exceptions were for fatty lesions, which have a pathognomonic appearance on both MRI and CT, and aggressive fibromatosis. In both of our cases with the latter condition, we observed low signal intensity from the tumor on MAI using both Ti - and T2-weighted pulse sequences. This may be characteristic for some of these lesions, when they consist of dense collagen and are relatively acellular. A major reason for the overall failure of MRI and CT in diagnosis is that calcification, ossification, and periosteal reaction are less easily evaluated than on plain radiographs. Calcification and cortical bone produce no signal on MRI, although identification is still often possible by recognizing the signal void. The findings are seen on CT, but with less spatial resolution than on routine radiographs. In several of our cases, the ability to visualize calcium did give CT a diagnostic advantage over MRI; in these cases, however, the plain films provided most of the essential missing information. It must be remembered, however, that much of the diagnostic value of plain films stems from well defined diagnostic criteria, which evolved from many years experience. Possibly, in this context,

AJR:146, April 1986 MRI/CT OF BONE & SOFT-TISSUE TUMORS 755 Fig. 7-22-year-old woman with parosteal osteosarcoma of distal left femur. A, Non-contrast-enhanced CT demonstrates lesion; bone-related artifact is present in marrow space. B, Transverse SE 2000/28 image at same level; sharp boundary of marrow fat is more easily appreciated. C, Sagittal SE 2000/28 image shows extent of cortical lesion (arrow).0, Plain film. TABLE 2: Calculated Relaxation Times Tl,msec (mean±5d) T2,msec (mean±sd) Normal muscle 577 ± 209 30 ± 4 Normal fat 345 ± 133 50± 7 Lesions 1054 ± 692 62 ± 18 the utility of both MRI and CT, especially MRI, will improve in the future. Calculated Ti and T2 relaxation times were not found to be useful in evaluating these lesions, because of both the wide variation in the relaxation times of tumors and technical difficulties in making the measurements with current equipment. The latter include partial-volume averaging effects, antenna-related radiofrequency field inhomogeneity, motion artifacts, the limited number of data points, and the approximations inherent in the algorithms used. Further, the small number of patients with each tumor type precluded the collection of statistically useful data. As more patients are studied and the technology is improved, such measurements may become practical and important. We did not recognize any parameters for differentiating benign from malignant lesions on the basis of relaxation time determinations or other characteristics unique to MRI. Nonetheless, we believe accepted criteria developed for other imaging methods, such as degree of invasiveness, sharpness of lesion boundary, and tumor homogeneity, are well suited to evaluation on MAI with its intrinsic high tissue contrast. Also, we have noted a sharp low-signal boundary in some benign lesions, including a posttraumatic fracture with a hematoma not included in this series (unpublished observation). A similar finding in bone marrow was described by Zimmer et al. [4]. We are currently evaluating the prevalence of such boundaries in benign osseous lesions and whether they have value in differential diagnosis. A more definite evaluation of the accuracy of MRI in defining the boundaries of actual neoplasm and in assessing the response to therapy must await additional studies with good radiologic-pathologic correlation. Nonetheless, we believe on the basis of our limited evidence that MRI, interpreted in

756 AISEN ET AL. AJR:146, April 1986 conjunction with appropriate conventional radiographs, is currently the method of choice in assessing the extent of disease for tumors of the skeletal and somatic soft tissues. In most cases, CT will not be a necessary additional test, though in a minority of patients it may provide important complementary information, because of its ability to demonstrate calcium. REFERENCES 1. Brady TJ, Rosen BR, Pykett IL, McGuire MH, Mankin HJ, Rosenthal DI. NMR imaging of leg tumors. Radiology 1983;i49:i81-i87 2. Moon KL, Genant HK, Helms CA, Chafetz NI, Crooks LE, Kaufman L. Musculoskeletal applications of nuclear magnetic resonance. Radiology 1983;i 61:161-171 3. Hudson TM, Hamlin DJ, Enneking WF, Petterson H. Magnetic resonance imaging of bone and soft tissue tumors: early experience in 31 patients compared with computed tomography. Skeletal Radiol 1985;1 3:134-146 4. Zimmer WD, Berquist TH, McLeod RA, et al. Bone tumors: magnetic resonance imaging versus computed tomography. Radiology 1985;i 55:709-718