Comparison of single photon emission computed tomography-computed tomography, computed tomography, single photon emission computed tomography and planar scintigraphy for characterization of isolated skull lesions seen on bone scintigraphy in cancer patients.

PURPOSE: The purpose of this study is to evaluate the added value of single photon emission computed tomography-computed tomography (SPECT-CT) over planar scintigraphy, SPECT and CT alone for characterization of isolated skull lesions in bone scintigraphy (BS) in cancer patients.
MATERIALS AND METHODS: A total of 32 cancer patients (age: 39.5 ± 21.9; male: female - 1:1) with 36 isolated skull lesions on planar BS, underwent SPECT-CT of skull. Planar BS, SPECT, CT and SPECT-CT images were evaluated in separate sessions to minimize recall bias. A scoring scale of 1-5 was used, where 1 is definitely metastatic, 2 is probably metastatic, 3 is indeterminate, 4 is probably benign and 5 is definitely benign. With receiver operating characteristic analysis area under the curves (AUC) was calculated for each modality. For calculation of sensitivity, specificity and predictive values a Score ≤3 was taken as metastatic. Clinical/imaging follow-up and/or histopathology were taken as reference standard.
RESULTS: Of 36 skull lesions 11 lesions each were on frontal, parietal and occipital bone while three lesions were in the temporal bone. Of these 36 lesions, 16 were indeterminate (Score-3) on planar and SPECT, five on CT and none on SPECT-CT. The AUC was largest for SPECT-CT followed by CT, SPECT and planar scintigraphy, respectively. Planar scintigraphy was inferior to SPECT-CT (P = 0.006) and CT (P = 0.012) but not SPECT (P = 0.975). SPECT was also inferior to SPECT-CT (P = 0.007) and CT (P = 0.015). Although no significant difference was found between SPECT-CT and CT (P = 0.469), the former was more specific (100% vs. 94%).
CONCLUSION: SPECT-CT is better than planar scintigraphy and SPECT alone for correctly characterizing isolated skull lesions on BS in cancer patients. It is more specific than CT, but provides no significant advantage over CT alone for this purpose.

INTRODUCTION

Bone is one of the most common sites of distant metastasis in cancer patients apart from lung and liver.[1] Various anatomical and functional imaging modalities are used for detection and characterization of metastasis. Among them, bone scintigraphy (BS) is a widely used procedure. It provides a whole-body skeletal survey at a relatively low cost and is usually the initial imaging modality for assessment of bone metastases.[2] Numerous reports emphasize the high sensitivity of BS in the diagnosis of osseous metastases. However, it lacks specificity due to metabolic reaction of bone to a variety of disease processes, including trauma and inflammation.[3] Skull bones are a common site for bone metastasis.[4] Characterization of isolated lesions in the skull seen on BS may represent a challenge because it is a flat bone and is a frequent site of traumatic lesions. Single-photon emission computed tomography (SPECT) improves the lesion-to-background contrast and sensitivity of 99m technetium methylene diphosphonate (99mTc-MDP) BS.[5] It enables accurate localization of tracer activity, especially in complex skeletal structures, such as skull and therefore can improve diagnostic specificity.[5] However, the specificity of SPECT is also not sufficient for a reliable diagnosis.[6] Correlation with high-quality anatomic images, computed tomography (CT), or magnetic resonance imaging (MRI), may still be needed. Unfortunately, recently done anatomic images may not be available at the time of the nuclear medicine procedure. Although, co-registration of anatomical and functional data obtained separately with different devices has been attempted using external fiducial markers, errors may occur as a result of variations in patient positioning.[7]

Recently state of the art hybrid SPECT-CT systems have become available which combine both tomographic scintigraphy and CT, producing a unique combination of functional and anatomical set of data.[8] These systems allow the field of view of the CT scan to be adapted to the SPECT findings. SPECT-CT has been shown to be useful for various indications and for different regions.[91011] However, until date only one study has systematically evaluated the efficacy of 99mTc-MDP hybrid SPECT-CT for skull lesions.[12] In addition, no study has compared SPECT-CT with SPECT and CT alone, which are more widely available. Therefore the aim of the present study was to compare the roles of planar scintigraphy, SPECT, CT and SPECT-CT for characterization of isolated skull lesions seen on BS.

MATERIALS AND METHODS

Patients

This was a retrospective analysis and was approved by the institutional review board. Between July 2009 and May 2012 a total of 52 patients with underlying malignancy showed isolated skull lesions (≤2 lesions/patient) on BS. Of these 32 patients had undergone additional SPECT and SPECT-CT. Data of these 32 patients was analyzed. As similar pattern of involvement of multiple skull lesions usually suggest a particular etiology, patients with >2 skull lesions were not included.

Radiotracer injection and planar scintigraphy

The patients were intravenously injected 666-925 MBq (18-25 mCi) of 99mTc-MDP, depending on the body weight. Planar scintigraphy was performed 3 h after radiotracer injection. Planar images were acquired either on a dual head gamma camera (Symbia E, Seimens Medical Solutions, Illinois, USA) or hybrid SPECT-CT dual-head gamma camera (Symbia T6, Seimens Medical Solutions, Illinois, USA). Anterior and posterior whole body planar images were acquired in a continuous mode by use of parallel-hole, low-energy, high-resolution collimators, with the patient in the supine position. Images were acquired on the 140-keV photopeak with a 20% symmetrical window and matrix size was 256 × 1024. Immediately after acquisition, the planar images were evaluated by a nuclear medicine physician regarding the additional imaging in the form of SPECT and SPECT-CT. For patients with isolated ≤ 2 skull (calvarial) lesions SPECT-CT was performed.

SPECT acquisition

SPECT of the skull was acquired using a hybrid SPECT-CT dual-head gamma camera (Symbia T6, Seimens Medical Solutions, Illinois, USA). Emission data were acquired by use of parallel-hole, low-energy, high-resolution collimators, with the patient in the supine position. The acquisition orbits were body contour orbits over 360° arcs, with the use of 60 stops each of 6°. For 60 stops, emission data were acquired for 30 s/stop. The image acquisition matrix was 128 × 128 and the pixel size was 4.8 mm. Images were acquired on the 140-keV photopeak with a 20% symmetrical window.

CT acquisition

The SPECT was followed by CT examination with acquisition parameters of 130 kV, 100 mAs, pitch-1, 512 × 512 matrix using standard filters. The CT images were reconstructed with reconstruction with B08 kernel for attenuation correction and B60 kernel for bone imaging.

Processing of SPECT images and co-registration

All studies were uniformly processed with commercially available E.soft (Seimens Medical Solutions, Knoxville, TN, USA) software on a Syngo nuclear medicine workstation (Seimens Medical Solutions, Illinois, USA). SPECT emission image data was processed by use of ordered-subsets expectation maximization reconstruction software with two iterations and eight subsets. A Guassian filter with full width at half maximum of 7.0 was applied. Attenuation correction was applied to these images using the CT based attenuation maps. The attenuation maps were created from the input CT image by converting the CT numbers to attenuation numbers, using look-up table, based on both CT effective energy spectrum (kVeff) and the emission isotope energy. Scatter correction was also applied. The corrected SPECT images were again reconstructed with Flash-3D software (Seimens Medical Solutions, Knoxville, TN, USA) with eight subsets and eight iterations. Subsequently, tomographic slices were generated and displayed as transaxial, coronal and sagittal slices. SPECT emission images were co-registered and fused with the transmission CT images using object versus target matrix method. Fused emission and transmission images were visually inspected for correctness of co-registration. Studies with significant misregistration were excluded from further analysis.

Image analysis

Planar, SPECT, CT and SPECT-CT images were analyzed by two experienced nuclear medicine physician with experience of SPECT-CT. The reader was blinded to patient's clinical information including diagnosis and findings of other imaging modalities, if any. Planar, SPECT, CT and SPECT-CT images were evaluated in separate sessions 1 week apart to minimize recall bias. The images were displayed in a random order. Only the lesions seen on planar scintigraphy were evaluated. In case of any discrepancy regarding findings of planar and SPECT images a consensus was reached after mutual discussion. On CT, malignant lesions were suggested by the presence of lytic, sclerotic, or mixed lytic-sclerotic changes. If there was any discrepancy regarding CT and SPECT-CT findings, the opinion of an experienced radiologist (NAF) was sought. Planar scintigraphy, SPECT, CT and SPECT-CT were compared in terms of the number of equivocal findings and accuracy on a lesion by lesion basis. The site and nature of lesions were also noted.

Receiver operating characteristic curve analysis

For the purpose of constructing receiver operating characteristic (ROC) curves, the interpreters used a scoring scale of 1-5, in which 1 is definitely metastatic, 2 is probably metastatic, 3 is indeterminate, 4 is probably benign and 5 is definitely benign. For the calculation of sensitivity, specificity and predictive values for planar scintigraphy, SPECT and SPECT-CT an interpretive Score ≤3 was taken as metastatic and with Score ≥4 was taken as benign.

Assessment of CT dose

For each patient, the dose parameters volume-weighted CT dose index (CTDIvol) and dose length product (DLP) were available in the patient protocol and were recorded. DLP is the product of CTDIvol (mGy) and scan length (cm). The DLP (mGy.cm) was then multi-plied with the appropriate conversion factor depending on the region of the body scanned to yield the effective dose (mSv) due to additional CT.

Reference standard

Final diagnoses (presence or absence of the skull bone metastases) were derived from clinical/imaging follow-up (CT, MRI, radiography, positron emission tomography (PET)-CT, SPECT-CT) over at least 6 months and/or histopathology (when available). Increase in size or change of character (lytic to sclerotic) under therapy was considered as positive for tumor, whereas lesions with unchanged size and character over 6 months without treatment were regarded as benign.

Statistical analysis

We expressed continuous data as mean ± standard deviation while categorical data was expressed as the number and percentage. For quantitative interpretation of the ROC curves, the area under the curve (AUC) was calculated and compared. A larger area indicates improved diagnostic performance. Sensitivity, specificity and predictive values, were separately calculated for planar scintigraphy, SPECT and SPECT-CT taking a Score of ≤3 as malignant. All statistical analysis was performed using Statistical Package for the Social Sciences 11.5 (SPSS Inc., Illinois, USA) and STATA (STATA Corp., College Station, Texas, USA).

RESULTS

Patient demographics including age, sex and indication of skeletal scintigraphy are detailed in Table 1. A total of 36 lesions were evaluated in these 32 patients. The site of these lesions is detailed in Table 2. The additional radiation exposure due to CT was 0.5 ± 0.7 mSv (range: 0.3-1.5).

Table 1

Patient characteristics

Table 2

Site of lesions

Based on the reference standard mentioned above, 55.5% (20/36) lesions were metastatic while 44.5% (16/36) lesions were benign. For five lesions, osteolysis and bone destruction were so obvious on SPECT-CT images that they were referred immediately for further treatment. Follow-up for validation was considered unnecessary in these patients. For remaining lesions, final diagnoses were derived from biopsies in three lesions and imaging follow-up over at least 6 (range: 6-12) months (CT, MRI, radiography, PET-CT, SPECT-CT) for 28 lesions. Detail of individual patient including follow-up is shown in Table 3.

Table 3

Patient wise summary of clinical, imaging and follow-up details

Planar scintigraphy, SPECT, CT and SPECT-CT

On planar scintigraphy 16 lesions were indeterminate and on SPECT 16 lesions were indeterminate. Five lesions were indeterminate on CT and none on SPECT-CT. Score of lesions on each modality is detailed in Table 4. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy of planar scintigraphy, SPECT and SPECT-CT are detailed in Table 5. SPECT-CT and CT were especially helpful for lytic lesions (n = 7) as compared to SPECT and planar BS.

Table 4

Score of lesions on different modalities

Table 5

Sensitivity, specificity, PPV, NPV and accuracy of planar scintigraphy, SPECT, CT, and SPECT-CT (with 95% confidence interval)

ROC analysis

The results of ROC analysis is shown in Table 6. The AUC was largest for SPECT-CT followed by CT, SPECT and planar scintigraphy respectively. We compared the diagnostic accuracy of planar scintigraphy, SPECT, CT and SPECT-CT by comparing the AUC [Figure 1]. The diagnostic accuracy of SPECT was not significantly different from planar scintigraphy (P = 0.975). SPECT-CT performed better than both planar scintigraphy (P = 0.006) and SPECT alone (P = 0.007), but was not superior to CT (P = 0.469). CT was also superior to planar scintigraphy (P = 0.012) and SPECT (P = 0.015).

Table 6

Results of receiver operating characteristic analysis for planar scintigraphy, SPECT, CT, and SPECT-CT

Figure 1

Receiver operating characteristic curve showing the area under the curve (AUC) for planar scintigraphy, single photon emission tomography (SPECT), computed tomography (CT) and SPECT-CT. the AUC was largest for SPECT-CT followed by CT, SPECT and planar scintigraphy, respectively

Incremental value of SPECT-CT

As these skull lesions were the only lesions in these patients, their management was dependent on characterization of these lesions. SPECT-CT correctly characterized 94% (15/16) of equivocal lesions seen on planar scintigraphy. In addition, eight definitely metastatic/probably metastatic lesions on planar scintigraphy were correctly characterized as benign on SPECT-CT. SPECT-CT correctly characterized 94% (15/16) indeterminate lesions seen on SPECT. Furthermore four definitely metastatic/probably metastatic lesions on SPECT were correctly characterized as benign on SPECT-CT. SPECT-CT also correctly characterized all 100% (5/5) equivocal lesions seen on CT. Representative images are presented in Figures 2 and 3.

Figure 2

A 24-year-old female with locally advanced carcinoma breast. Bone scintigraphy (BS) was done for restaging. Planar BS images (a and b) Focal uptake frontal bone almost in midline (arrow; Score 3). Axial single photon emission tomography (SPECT) (c) Uptake in the frontal bone just to the left of midline (arrow; Score 1). Axial computed tomography (CT) (d) and SPECT-CT (e) Increased tracer uptake in the left frontal sinus with no corresponding CT abnormality (arrow; both Score 5). In this case CT and SPECT-CT characterized the planar scintigraphy indeterminate lesion as benign while SPECT characterized it as metastasis. On follow-up BS with SPECT-CT the lesion disappeared without any systemic treatment confirming benign nature

Figure 3

A 25-year-old female with locally advanced breast cancer. Bone scintigraphy (BS) was done for staging. Planar BS images (a and b) Focal uptake in left frontal bone (arrow; Score 3). Axial single photon emission tomography (SPECT) (c) image shows uptake in the frontal bone (arrow; Score 3). Axial computed tomography (CT) (d) Minimal lytic changes in the left frontal bone (arrow; Score 2). SPECT-CT (e) Image reveals increased tracer uptake in the left frontal lesion (arrow; Score 1). SPECT-CT correctly characterized the lesion as metastasis and improved the diagnostic confidence over CT. on follow-up positron emission tomography-CT the lesion showed intense 18 fluoride-fluorodeoxyglucose uptake confirming metastatic nature

DISCUSSION

Skull lesions on BS are detected in many patients with underlying malignancy and can be due to a wide variety of benign and malignant diseases. The specificity of planar BS for characterization of skull lesions is limited. In the present study planar BS showed specificity of only 6.25%. In addition 16 lesions (44%) remained indeterminate on planar BS. When such skull lesions are the only lesions, as in our patient population, management of the patients depends on accurate characterization of these lesions. Hence, it is crucial to further evaluate such lesions. Addition of SPECT is known to improve the diagnostic accuracy of planar BS. Moreover, using SPECT alone does not entail any additional radiation exposure to the patient, apart from that due to 99mTc-MDP administration. However for skull lesions the specificity of SPECT alone remains low. This is due to the fact that localization alone does not help much for these lesions and anatomical correlation is usually required. The specificity of SPECT in the presents study was 12.5% with 44% lesions remaining indeterminate. In fact, there was no significant difference between BS and SPECT for characterization of skull lesions (P = 0.975). SPECT-CT combines the functional information of SPECT with anatomical information of CT. Römer et al. first evaluated the role of SPECT-CT for characterizing indeterminate bony lesions in patients with malignancy.[13] SPECT-CT was able to clarify 90% such lesions in cancer patients. Sharma et al. evaluated SPECT-CT for characterization of vertebral lesions seen on planar BS.[11] SPECT-CT was found to be superior to planar scintigraphy and borderline superior to SPECT but not CT. They concluded that SPECT-CT can have a significant impact of patient management.

Gayed et al. evaluated SPECT-CT for characterization of solitary skull lesion seen on BS in 19 patients.[12] In their study 71% of the lesions were correctly characterized by SPECT-CT.

The sensitivity, specificity, PPV and NPV of SPECT-CT was 100%, 92%, 75% and 100%, respectively. They concluded that SPECT-CT could correctly characterize such lesions. In the present study we found similar results with sensitivity, specificity, PPV and NPV being 95%, 100%, 100% and 94%, respectively. SPECT-CT was superior to planar scintigraphy (P = 0.006). It characterized 94% of equivocal lesion seen on planar scintigraphy and characterized 44% of metastatic/probably metastatic lesions seen on planar scintigraphy as benign. SPECT-CT was also superior to SPECT alone (P = 0.006).

None of the studies in the literature have compared CT alone with SPECT-CT for characterization of skull lesions seen on BS. We evaluated CT in this scenario and found it to be superior to planar scintigraphy and SPECT. It was also not inferior to SPECT-CT (P = 0.469). However even on CT five lesions still remained indeterminate. The specificity of CT (94%) was also lower than SPECT-CT (100%). The radiation exposure due to additional CT was also low (0.5 ± 0.7 mSv), which is much lower compared to that due to BS alone (3-4 mSv).[14] Given the added advantage of SPECT-CT and low additional radiation burden, SPECT-CT should be routinely employed for characterization of skull lesions seen on planar BS.

The present study had certain limitations. Firstly, this was a retrospective analysis. Secondly, histopathological diagnosis was not available for all lesions and imaging was the mainstay of confirming the diagnosis. Though this is not ideal, it is acceptable given the difficulties and ethical issues associated with bone biopsy. Further prospective studies addressing these shortcomings and comparing SPECT-CT with other modalities such as 18-fluorodeoxyglucose PET-CT and 18-fluoride PET-CT are warranted.

CONCLUSION

SPECT-CT is superior to planar BS and SPECT for characterization of isolated skull lesion seen on 99mTc-MDP BS. It is more specific than CT but provides no significant advantage over CT alone for this purpose.