Screening for bone metastases: whole-body MRI using a 32-channel system versus dual-modality PET-CT
- First Online:
- Cite this article as:
- Schmidt, G.P., Schoenberg, S.O., Schmid, R. et al. Eur Radiol (2007) 17: 939. doi:10.1007/s00330-006-0361-8
- 644 Views
The diagnostic accuracy of screening for bone metastases was evaluated using whole-body magnetic resonance imaging (WB-MRI) compared with combined fluorodeoxyglucose (FDG) positron emission tomography (PET) and computed tomography (CT) (FDG-PET-CT). In a prospective, blinded study, 30 consecutive patients (18 female, 12 male; 24–76 years) with different oncological diseases and suspected skeletal metastases underwent FDG-PET-CT as well as WB-MRI with the use of parallel imaging (PAT). With a 32-channel scanner, coronal imaging of the entire body and sagittal imaging of the complete spine was performed using T1-weighted and short tau inversion recovery (STIR) sequences in combination. PET-CT was conducted using a low-dose CT for attenuation correction, a PET-emission scan and diagnostic contrast-enhanced CT scan covering the thorax, abdomen and pelvis. Two radiologists read the MRI scans, another radiologist in combination with a nuclear medicine physician read the PET-CT scans, each in consensus. The standard of reference was constituted by radiological follow-up within at least 6 months. In 28 patients, 102 malignant and 25 benign bone lesions were detected and confirmed. WB-MRI showed a sensitivity of 94% (96/102), PET-CT exams achieved 78% (79/102; P<0.001). Specificities were 76% (19/25) for WB-MRI and 80% (20/25) for PET-CT (P>0.05). Diagnostic accuracy was 91% (115/127) and 78% (99/127; P<0.001), respectively. Cut-off size for the detection of malignant bone lesions was 2 mm for WB-MRI and 5 mm for PET-CT. WB-MRI revealed ten additional bone metastases due to the larger field of view. In conclusion, WB-MRI and FDG-PET-CT are robust imaging modalities for a systemic screening for metastatic bone disease. PAT allows WB-MRI bone marrow screening at high spatial resolution and with a diagnostic accuracy superior to PET-CT.
KeywordsBoneMetastasesMagnetic resonanceComputed tomographyPositron emission tomography
The skeletal system is a frequent target of metastatic disease and early detection of bone metastases has an important impact on patient management, disease outcome and the quality of life of the patient . In clinical practice, multimodality algorithms are widely applied in case of suspected metastatic bone disease, including conventional X-ray, skeletal scintigraphy, positron emission tomography (PET), computed tomography (CT) and magnetic resonance imaging (MRI). Plain radiographs have a low sensitivity for the detection of bone metastases and only become apparent after a loss of more than 50% of bone mineral content . At present, 99mTc-phosphonate-based skeletal scintigraphy is the standard method for initial staging. However, at an early stage of disease, lesions may remain invisible in the absence of an osteoblastic response . Furthermore, false-positive findings may arise by a misinterpretation of tracer uptake in healing fractures or degenerative disease. Recent studies have indicated that whole-body fluorodeoxyglucose (FDG)-PET increases the specificity of bone marrow screening compared with scintigraphy, due to tracer uptake directly into malignant cells [3, 4]. Fused PET-CT scanners combine the functional data of PET with the anatomical information of CT scanners in a single examination and have further improved diagnostic accuracy and lesion localization [5, 6]. Moreover, the CT image data allow assessment of paraosseous tumor expansion and provide information on the extent of osteolysis as well as criteria of bone stability.
In contrast, MRI is an imaging technique that provides visualization of the bone marrow components at a high spatial resolution and has proven to be very sensitive for the early detection of bone marrow pathologies [7–9]. The fact that 40% of skeletal metastases occur in the appendicular skeleton stresses the need for accurate bone marrow imaging covering the whole body anatomy . However, different requirements in coil setup, sequence design and slice positioning, as well as time-consuming patient repositioning procedures in the past, have delayed the realization of whole-body MR imaging as a clinical application. With the recent introduction of multi-channel whole-body scanners, covering the patient’s anatomy from head to toe, with its lack of ionizing radiation and excellent soft tissue contrast, MRI has become a promising candidate for whole-body bone marrow screening. Various studies have described the efficiency of MRI over CT and skeletal scintigraphy in the detection of primary bone neoplasms and metastases [3, 8, 10, 11]. The combination of non-contrasted T1-weighted spin-echo (SE) sequences and fat-suppressed short tau inversion recovery (STIR) imaging proved to be most accurate in the detection of malignant bone marrow disorders and an excellent negative and positive predictive value especially for STIR-imaging has been described in literature [9, 12–15]. Additionally, MRI enables precise assessment of the tumor extent within the bone marrow and into paraosseus structures, such as the spinal canal.
WB-MRI and FDG-PET-CT represent two imaging modalities that potentially can detect metastatic bone disease at an early stage of growth, before an osteoblastic host reaction occurs [7, 8, 16]. The purpose of this study was to compare the diagnostic potential of these whole-body modalities for the detection of bone metastases and to establish an MR protocol based on fast T1-weighted and STIR imaging, covering the entire body from head to toe.
Materials and methods
The study included 30 patients (18 female, 12 male, mean age 58 years, range 24–76 years) with different histologically clarified primary tumors or cancer of unknown primary (CUP). Primary tumor diagnoses included breast carcinoma (n=13), tumors of the gastrointestinal tract (n=7), CUP (n=4), malignant melanoma (n=3), hepatocellular carcinoma (n=1), non-Hodgkin lymphoma (n=1) and rhabdomyosarcoma (n=1). The patients were referred to our hospital within a time period of 5 months for staging and follow-up of known skeletal metastases (n=7), suspicion of bone metastases in another imaging modality (n=6) or diagnostic work-up of newly occurring bone pain and/or elevated tumor marker levels (n=17). All patients were examined consecutively with a dual-modality PET-CT-scanner (Gemini, Philips Medical Systems, Cleveland, Ohio) within the diagnostic algorithms of clinical routine. These patients additionally underwent WB-MRI (Magnetom Avanto, Siemens Medical Solutions, Erlangen, Germany) within a maximum of 14 days. MRI was performed in agreement with the patients as well as the clinicians in charge and by approval of the institutional review board. Written patient consent was obtained before both examinations. Both PET-CT and WB-MRI were well tolerated by all patients. To avoid bias in diagnostic accuracy on PET-CT due to suppressed metabolic activity, three patients receiving chemotherapy less than 1 month prior to our examinations were excluded from statistical analysis. No chemotherapy or radiation therapy was performed in between both examinations.
Magnetic resonance imaging
Examinations were performed on a dual-modality two-detector row PET-CT-scanner (Gemini, Philips Medical Systems, Cleveland, Ohio) after injection of 204–326 MBq [18F]-fluoro-2-deoxy-D-glucose. Patients were asked to fast for at least 6 h before examination to assure blood glucose levels below 150 mg/dl. Buscopan was applied intravenously to avoid a first-pass uptake of FDG into smooth muscle. Additionally, 20 mg of furosemide was given to increase renal excretion of the tracer and avoid accumulation in non-malignant cells. One hour after FDG administration, a low-dose CT scan in shallow breathing was performed for PET attenuation correction covering the neck, thorax, abdomen and pelvis (40 mAs, 120 kV, collimation 2×5 mm, pitch 1.5). Then the emission scan followed using an integrated GSO crystal-based PET system with 3D RAMLA (Row Action Maximum Likelihood algorithm) reconstruction (ten bed positions, FOV 10 cm, 144×144 matrix). Finally, the diagnostic contrast-enhanced CT was conducted (120 mAs, 160 kV, collimation 2×5 mm, pitch 1) with application of 120 ml of i.v. contrast agent (Ultravist® 300, Schering, Berlin, Germany) in the venous phase (80 s delay). Finally, the PET and diagnostic CT data were fused with the use of special software (Syntegra, Philips Medical Systems, Cleveland, Ohio). For the whole examination, an average dose of ionizing radiation of 25 mSV has been calculated . Examination time for PET-CT was 103 min (60 min patient preparation, total acquisition time 43 min).
Two board-certified radiologists, each with more than 6 years of experience, read the MRI scans and another radiologist and one nuclear radiologist with 3 and 6 years of experience read the PET-CT images, each in consensus. Both reader groups were fully blinded to the other modality and had no information on previous or current diagnostic imaging results. Solely information on the primary tumor diagnosis was provided. Location, extension, size and number of suspected malignant or benign lesions to the skeletal system were recorded. On MRI, a lesion was considered to be malignant when there was a focal or a diffuse hypointense bone marrow signal compared with adjacent muscle on T1-weighted SE images, with a corresponding intermediate or hyperintense signal on STIR imaging, which could not be explained by a degenerative, inflammatory or traumatic cause . To further differentiate metastases from benign lesions, additional criteria like the “bull’s eye” and “halo sign” were considered as described in published literature . Areas of low signal intensity on both T1-weighted and STIR sequences were interpreted as sclerosis. Furthermore, criteria for malignancy were paraosseous tumor infiltration or, for evaluation of the spine, signal changes extending into the pedicles and bulging of the anterior or posterior margin of the vertebral body. Criteria for the benign nature of a lesion were high signal intensity on T1-weighted SE images or a location near joint surfaces or degenerative changes of the vertebral endplates. Mild heterogenous bone marrow changes were considered to be inhomogenous fatty replacement of red marrow or recovering marrow. On PET-CT images, malignancy was defined by a focally increased FDG-uptake, unless explained by other conditions, like degenerative processes, adequate trauma or signs of infection. Also, measurements of the maximum “standard uptake value” (SUVmax) were taken in suspicious lesions. In uncertain cases, an SUVmax of more than 2.5 was considered indicative of a malignant process . Using morphological information in the diagnostic CT, malignant lesions were suggested by the presence of lytic, sclerotic or mixed lytic-sclerotic intramedullary changes or bone lesions with accompanying adjacent soft tissue abnormality. In the spine, lesions located at the posterior part of the body or at the pedicles were considered to be malignant, whereas lesions located at the facet joints, end plates or at the posterior aspect of the spinous process were considered benign.
Lesion size was defined by the largest diameter and, in case of multifocal disease, representative smallest and largest lesions were measured. Focal masses were counted lesion-by-lesion. In case of a diffuse infiltration pattern an evaluation system of anatomical regions was used, covering the following areas: skull, cervical spine, shoulder girdle, rib cage, thoracic spine, lumbar spine, sacrum, pelvis, proximal upper extremity, distal upper extremity, proximal lower extremity and distal lower extremity. Each affected bone region was treated like one lesion. Thus, for example a diffuse infiltration of the cervical and thoracic spine was counted as 2 or an infiltration of both clavicles and left humeral bone in total as 3. All detected lesions, including all questionable or discordant findings between both modalities, were cross-checked by radiological follow-up studies within at least 6 months as a standard of reference. The following studies were performed: PET-CT=10, CT=8, bone scintigraphy=7, radiographs=5, MRI=5, WB-MRI=2. Metastatic disease was confirmed by an obvious progression in the number and/or size of lesions on follow-up examinations, by signs of progressive infiltrative growth of the lesion or by typical criteria of a malignant lesion on a follow-up examination other than MRI or PET-CT. Benign findings in this study included atypical hemangioma, enchondroma, osteochondroma, Schmorl’s nodes and focal islands of hematopoetic bone marrow.
To assess diagnostic accuracy for both modalities, sensitivity, specificity, and positive and negative predictive values were calculated for the observed bone lesions. Statistical differences in diagnostic performance between PET-CT and WB-MRI on a lesion-by-lesion basis were tested using the McNemar test. A P value of less than 0.05 was considered to indicate a statistically significant difference. Size measurements were reported as median value and range. Statistical analyses were performed with SPSS for Windows (Version 11.0, SPSS, Chicago, Ill.).
Altogether PET-CT and WB-MRI detected 126 of 127 bone lesions in 28 patients, two patients had no visible pathologies of the bone and also were negative on follow-up examinations. One lesion previously known in other staging procedures (dedicated MRI of the upper spine/conventional X-ray) was missed by both PET-CT and WB-MRI. The 126 detected lesions were observed within the “overlapping” field of view of both modalities from the skull base to the proximal femoral bone. Of these lesions, 102 were confirmed as malignant, 25 as benign. Findings were concordant for both modalities in 72% (91/127) of the detected bone lesions. All 36 discordant findings were unequivocally classified by follow-up examinations.
In a patient-by-patient analysis, metastatic bone disease was revealed in all these 23 patients by WB-MRI (sensitivity 100%), but was missed in two patients by PET-CT (sensitivity 91%). WB-MRI falsely diagnosed metastatic bone disease in one patient that showed no evidence indicating bone malignancy in follow-up (specificity 80%) and was confirmed as focal hematopoetic bone marrow. PET-CT showed a specificity of 100%. Patient-based diagnostic accuracy was 96% for WB-MRI and 93% for PET-CT.
Diagnostic performance of WB-MRI and PET-CT in the detection of 127 bone lesions determined by sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV) and accuracy
Total bone lesions confirmed
PET-CT indicated five and WB-MRI six false-positive lesions: one lesion was false-positive in both modalities, five benign lesions were only seen in WB-MRI and four lesions were false-positive solely in WB-MRI.
Considering the full field of view of a WB-MRI examination, covering the body from head to toe, a total of ten additional malignant lesions of the bone could be detected in five patients. Seven of these lesions were located in the distal femoral bones (see Fig. 3), two lesions in the tibial bones and one lesion in the mandible. No metastases were found in the skull.
Bone-related additional diagnoses made by WB-MRI and PET-CT in 13 patients
Necrosis of the femoral head
Accurate detection of bone metastases in staging and restaging of patients with a neoplastic disease is vital to assess therapeutic options and patient prognosis. Recently, multidetector multislice-CT has been proposed fo whole-body screening of the bone as an alternative method to skeletal scintigraphy . Still, for a long period of time scintigraphy has been the standard method in clinical routine for assessment of the entire skeletal system. Compared with MRI, a limited specificity and lower sensitivity of both scintigraphy and CT in the early detection of skeletal metastases has been reported [3, 8, 10, 26]. There is conflicting evidence on the fact if FDG-PET as a competing whole-body modality is more sensitive than bone scintigraphy, but the benefit of tracer uptake directly into active tumor cells seems to lead to an increase in specificity [27–29]. Fused PET-CT scanners have made a new modality available for whole-body imaging, combining the functional data of PET with the detailed anatomical information of CT scanners in a single examination . Different authors report a significant decrease of ambiguous lesions and an improvement of specificity for PET-CT compared with PET alone for the detection malignant disease [6, 30–32]. On the other hand, the recent introduction of whole-body MRI scanners with multiple phased array surface coils and receiver channels, combined with acceleration techniques for image acquisition like PAT have made flexible whole-body MRI screening of the bone feasible [15–17]. Since the bone is the third most common site of metastasis, it is of interest how these modalities perform as a new tool for tumor search.
On a patient-by-patient analysis, WB-MRI revealed metastatic bone disease in all cases, PET-CT showed a good sensitivity of 91%. These figures reflect published data for WB-MRI and PET and indicate no significant gain for PET-CT concerning patient-based sensitivity compared with PET alone [22, 25, 28]. However, PET-CT showed 100% patient-based specificity, whilst WB-MRI was false-positive in one patient with atypical haemangioma. This patient reliably was discriminated in PET-CT through the advantage of additional metabolic information.
The lesion-by-lesion analysis performed in this study demonstrates both PET-CT and WB-MRI as robust imaging modalities for screening of skeletal metastases. However, the overall diagnostic accuracy was significantly higher in WB-MRI (91%) than in PET-CT (78%). This advantage is mainly due to the superior sensitivity observed for WB-MRI (WB-MRI 94% vs PET-CT 78%). Similar observations were made by the study group of Antoch and co-workers, who examined 98 patients with different oncological diseases with both modalities for TNM-staging . Although results showed similar overall sensitivities for the detection of distant metastases (WB-MRI 93% vs PET-CT=94%), sensitivity was significantly higher for the assessment of skeletal metastases when using whole-body MRI (WB-MRI 85% vs 62%). Our results indicate that the increase in sensitivity of PET-CT for the diagnosis of bone metastases is minor compared with several studies performed with PET alone [21, 22, 34]. Interestingly though, 13% of bone metastases diagnosed by PET-CT in our study were PET-negative and final diagnosis was made with the diagnostic CT data alone. The anatomical information of CT is particularly helpful in revealing sclerotic metastases, which occur frequently in patients with breast cancer and have been reported to be less FDG-avid than osteolytic metastases or mixed-type lesions .
Our observations show a slight advantage in specificity of PET-CT (80%) over WB-MRI (76%). Again, this is in agreement with the results of Antoch et al. . As FDG is a non-tumor-specific tracer it may also accumulate, e.g. in the presence of inflammation and thus lead to false-positives. Here the detailed anatomical information provided by the spiral-CT with the possibility of exact lesion localization represents a problem-solving tool and certainly leads to a decrease of false-positive findings. However, our data do not provide information on the question of if there was a significant improvement of specificity in PET-CT compared with PET and CT alone, as an “intra”-modality comparison of diagnostic accuracy was not subject of this study.
Conflicting results have been reported by a group that performed WB-MRI for the detection of bone metastases in children and young adults and compared WB-MRI with FDG-PET alone and skeletal scintigraphy . Although WB-MRI was superior to skeletal scintigraphy, FDG-PET showed the highest sensitivity (WB-MRI 82% vs PET 90% vs scintigraphy 71%). Possible explanations are known diagnostic problems in MRI caused by the higher cellularity of normal bone marrow in this age group, yielding low signal intensity on T1-weighted and a hyperintense signal on fat-suppressed sequences. Also, it is important to mention that the MR examinations in 29 of the 39 patients were based on T1-weighted imaging alone.
The fact that 40% of skeletal metastases occur in the appendicular skeleton stresses the need for accurate bone marrow imaging covering the whole body anatomy . Our data confirm these observations by demonstrating that 30% of all detected metastases were located outside of the field of view of a routine MRI bone-screening protocol covering just the spine, pelvis and proximal thirds of the femoral bones. Due to the larger field of view, ten more metastatic lesions located outside the diagnostic range of PET-CT were revealed by WB-MRI. Similar additional findings have been reported by other authors using analogous imaging protocols, either with a sequential or moving table-top scanning technique [10, 12]. These findings can be of importance and therapeutic relevance as there have been various reports on bone metastases found solely in the appendicular skeleton (even though this was not the case in our study) [37, 38]. It has been referred that the traceability of bone metastases in WB-MRI is influenced by their anatomical localization: WB-MRI seems to have disadvantages in detecting metastases located in the skull and curved flat bones, especially in the thoracic cage where additional motion artefacts of respiration and pulsation impair image quality [9, 10, 12]. These observations are only partly supported by our data. WB-MRI performed better in five of nine possible anatomical sites and equally to PET-CT in three more anatomical regions, indicating a high overall performance, irrespective of lesion location. Lesions missed in WB-MRI were predominantly located in the spine; however, two lesions missed in WB-MRI were again located in the rib cage. This overall improvement of traceability may be a result of an optimized spatial resolution and coil geometry by using the matrix coil system in combination with PAT acceleration of image acquisition. In addition, PAT may reduce artefacts in the thoracic region by reducing scan time when breath-hold techniques are used.
The main impairment of a whole-body MRI bone screening in the past have been long examination times up to an hour when sequential scanning was performed or a scan-time reduction at the cost of significant restrictions in the sequence protocol or image quality. Our proposed protocol shows that the acceleration of image acquisition due to PAT enables flexible MRI scanning of the entire skeleton within 45 min scan time without a compromise in spatial resolution. However, coil setup and patient positioning add up to an average of 54 min room time per patient. Still, this amounts to only 50% of the room time that has to be estimated for a routine PET-CT, including patient preparation. The high in plane-resolution achieved by the 32-channel scanner is reflected by a distinctly lower cut-off size observed for malignant lesion detection compared with PET-CT (2 mm vs 5 mm). Especially smaller sized-lesions (<1 cm) in our study were detected with a higher sensitivity by WB-MRI (WB-MRI 88% vs PET-CT 56%). Additionally, the high tissue contrast in STIR imaging allows an excellent delineation of focal lesions and thus made a number of lesions visible which show no adequate FDG uptake and who are not visible as osteolysis in CT. Especially lesions below double the size of the spatial resolution of the PET-scanner (usually 6 mm) can lead to false-negative results.
Both modalities showed a moderate concordance of 72% meaning that a certain number of differing lesions were missed or detected in both modalities. This in fact indicates a complementary role of both modalities for bone screening and reflects the described technical and diagnostic problems within each modality. Numerous additional diagnoses related to the skeletal system were made in 13 patients. However, only in two patients these were tumor-related and in these patients presented as pathological fractures. Such diagnoses, which alter patient management, were reliably detected by both imaging modalities.
A limiting aspect of this study was the heterogenous population of primary tumors which all presented with different types of skeletal metastases and infiltration patterns. Especially in diffuse bone marrow infiltration MRI might have a diagnostic advantage over PET-CT as visible osteolytic changes might not yet have taken place and a diffuse FDG-uptake pattern might be harder to diagnose compared to a focal increase. However, it must be noted that all patients presenting with a diffuse infiltration pattern in our study were in an advanced stage of disease with noticeable bone marrow destruction. On the other hand, the presence of small osteoblastic metastases can be difficult to diagnose in MRI and might favour CT, as previously discussed. Certain tumor types with frequent osseous metastatic disease, like prostate and lung cancer, are not represented in our population as patient inclusion was performed consecutively to avoid selection bias. Thus, the large proportion of patients suffering from breast cancer and tumors of the GIT-tract represent the typical spectrum of neoplasms examined at our institute. Finally, the lack of a histological proof as a true “gold standard” for the detected lesions and verification based on radiological follow-up alone represent another limiting aspect. With a reference standard based on imaging alone, e.g. false-negatives may arise in small or slowly growing lesions in absence of significant morphological changes. On the other hand, comparable to numerous studies of similar design, obtaining multiple biopsies for tissue verification would have been impracticable and ethically unacceptable [9, 10, 31, 32]. Altogether it is important to take into account that the diagnostic accuracy of both imaging modalities can be considerably influenced by several conditions. FDG-uptake is susceptible to chemotherapy and also dependent on the primary tumor type. On the other hand, avital lesions after successful chemotherapy can remain virtually unchanged in morphology or signal in MRI which may complicate evaluation of therapy response. To avoid bias, we excluded all patients who had prior systemic chemotherapy. The primary tumors represented in our patient population are all known to be FDG-avid, which might have favoured PET-CT in comparison to an even more heterogenous population.
It is important to stress that in near future these imaging modalities certainly will still be restricted to a university hospital setting or to specialized imaging centers. To reduce costs of these expensive techniques, it may be considered that the appendicular skeleton may only be examined in patients where the detection of metastases in these regions may have an impact on patient management. However, it has to be considered that through the quick systemic information acquired, hospitalization potentially could be reduced and costly, fruitless diagnostics, which can occur in multi-modality approaches, potentially be avoided.
In summary, the robust performances of WB-MRI and PET-CT may significantly improve the detection of skeletal metastases in various malignant diseases and represent a promising alternative to multi-modality diagnostic algorithms. With the use of PAT acceleration on a multi-channel whole-body MR scanner, finally bone marrow screening of the entire body with high resolution and precise diagnostic accuracy is feasible.