Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.
Purpose:
To assess the diagnostic performance of whole-body non– contrast material–enhanced positron emission tomography (PET)/magnetic resonance (MR) imaging and PET/computed tomography (CT) for staging and restaging of cancers and provide guidance for modality and sequence selection.
Materials and Methods:
This study was approved by the institutional review board and national government authorities. One hundred six consecutive patients (median age, 68 years; 46 female and 60 male patients) referred for staging or restaging of oncologic malignancies underwent whole-body imaging with a sequential trimodality PET/CT/MR system. The MR protocol included short inversion time inversion-recovery (STIR), Dixon-type liver accelerated volume acquisition (LAVA; GE Healthcare, Waukesha, Wis), and respiratory-gated periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER; GE Healthcare) sequences. Primary tumors (n = 43), local lymph node metastases (n = 74), and distant metastases (n = 66) were evaluated for conspicuity (scored 0–4), artifacts (scored 0–2), and reader confidence on PET/CT and PET/MR images. Subanalysis for lung lesions (n = 46) was also performed. Relevant incidental findings with both modalities were compared. Interreader agreement was analyzed with intraclass correlation coefficients and k statistics. Lesion conspicuity, image artifacts, and incidental findings were analyzed with nonparametric tests.
Results:
Primary tumors were less conspicuous on STIR (3.08, P = .016) and LAVA (2.64, P = .002) images than on CT images (3.49), while findings with the PROPELLER sequence (3.70, P = .436) were comparable to those at CT. In distant metastases, the PROPELLER sequence (3.84) yielded better results than CT (2.88, P , .001). Subanalysis for lung lesions yielded similar results (primary lung tumors: CT, 3.71; STIR, 3.32 [P = .014]; LAVA, 2.52 [P = .002]; PROPELLER, 3.64 [P = .546]). Readers classified lesions more confidently with PET/ MR than PET/CT. However, PET/CT showed more incidental findings than PET/MR (P = .039), especially in the lung (P , .001). MR images had more artifacts than CT images.
Conclusion:
PET/MR performs comparably to PET/CT in whole-body oncology and neoplastic lung disease, with the use of appropriate sequences. Further studies are needed to define regionalized PET/MR protocols with sequences tailored to specific tumor entities.
1
From the Department of Medical Radiology, Divisions of Nuclear Medicine (M.W.H., P.A., F.P.K., L.H., C.M.P., I.A.B., M.P., G.D., G.K.v.S., P.V.H.), Diagnostic and Interventional Radiology (P.A., F.P.K., P.V.H.), and Neuroradiology (M.W.H., F.P.K.), University Hospital Zurich, Rämistrasse 100, 8091 Zurich, Switzerland; and GE Healthcare, Waukesha, Wis (G.D.). Received January 23, 2014; revision requested March 17; revision received April 22; accepted June 10; final version accepted June 18. Address correspondence to M.W.H. (e-mail: [email protected]). RSNA, 2014
q
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
q RSNA, 2014
Online supplemental material is available for this article. 859
Medicine
Martin W. Huellner, MD Philippe Appenzeller, MD Félix P. Kuhn, MD Lars Husmann, MD Carsten M. Pietsch, MD Irene A. Burger, MD Miguel Porto, MRT Gaspar Delso, PhD Gustav K. von Schulthess, MD, PhD Patrick Veit-Haibach, MD
Original Research n Nuclear
Whole-Body Nonenhanced PET/ MR versus PET/CT in the Staging and Restaging of Cancers: Preliminary Observations1
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
T
o consider positron emission tomography (PET)/magnetic resonance (MR) imaging a clinical alternative to PET/computed tomography (CT), two major goals have to be achieved. First, efficient imaging protocols have to be developed, preferably not exceeding the examination time needed for PET/ CT. Second, the sensitivity and specificity of PET/MR and PET/CT should be comparable. Only then can PET/MR be considered a clinically viable alternative to PET/CT. This implies that complete MR data acquisition is performed during PET acquisition, which is approximately 10–20 minutes, or 2–4 minutes per bed position. Therefore, the key in clinical whole-body PET/MR must be the determination of MR protocols and the selection of adequate MR pulse sequences to provide information that is complementary and/or confirmatory to that of PET, rather than redundant. In contrast to CT, where protocol issues are essentially binary (low dose or high dose; contrast material enhanced or nonenhanced), in MR imaging, multiple pulse sequences can be chosen. Clinical studies are needed to compare PET/ CT with PET/MR to identify MR pulse sequences that are most complementary
Advances in Knowledge nn Our preliminary study suggests that PET/MR performs comparably to PET/CT in whole-body oncology (primary tumors, P = .436) and neoplastic lung disease (primary lung tumors, P = .546). nn PET/MR with a combination of a fast whole-body Dixon-type sequence and a regionalized (lung and upper abdomen) highresolution, respiratory-gated T2-weighted sequence might provide a slight advantage over PET/ CT by yielding a higher conspicuity of distant metastases. nn PET/MR exhibits more nonproblematic image artifacts (primary tumors, P = .029), while PET/CT leads to more incidental nononcologic findings (P = .039), especially in the lung (P , .001). 860
Huellner et al
to the PET examination and ideally not more time consuming than PET. Appenzeller et al demonstrated that MR/PET imaging with a body coil is inferior to PET/CT in oncology patients (1). Drzezga and coworkers reported no difference in lesion detection between PET/CT and PET/MR in 32 patients (2). However, assessment of lesion conspicuity and image-degrading artifacts was not part of their study. Whole-body MR was reported to have an accuracy similar to that of fluorine 18 fluorodeoxyglucose (FDG) PET/CT for non–small cell lung cancer T and N staging, with MR being superior in the detection of brain and liver metastases, while PET/ CT performed better in lymph node and soft-tissue metastases (3). Apparently, three-dimensional Dixon-based gradient-recalled-echo sequences, often used for MR-based attenuation correction, perform comparably to low-doseCT. A study performed with a sequential trimodality system showed that such a three-dimensional pulse sequence may be sufficient for lung imaging with PET/ MR (4). A similar breath-hold sequence was also found to be comparable to lowdose CT for the localization of oncologic lesions that were positive at PET in PET/ CT and coregistered MR (5). The performance of different nonenhanced MR pulse sequences versus nonenhanced CT with regard to lesion conspicuity and image-degrading artifacts has so far not been studied in a larger cohort. The purpose of this prospective study was to assess the diagnostic performance of whole-body nonenhanced PET/MR and PET/CT for staging and restaging of cancers to provide guidance for modality and sequence selection.
Materials and Methods This study was approved by the institutional review board and by national government authorities. All subjects
Implication for Patient Care nn In oncology patients, cancers can be effectively staged and restaged by using PET/MR instead of PET/ CT.
provided signed informed consent prior to the examinations. Financial support was provided by an institutional grant from GE Healthcare. Only non-GE employees had control of inclusion of data and information that might present a conflict of interest for authors who are employees of GE Healthcare. In this prospective study, 106 consecutive patients (46 female and 60 male patients; median age, 68 years; range, 16–87 years) referred for staging (n = 26) or restaging (n = 80) of different oncologic malignancies between April 2012 and October 2012 underwent imaging with a PET/CT/MR system (detailed information is provided in Appendix E1 [online]).
Reference Standard The standard of reference consisted of histologic findings, if available, and clinical and imaging follow-up (mean, 335 days; range, 290–512 days). Histologic findings were available for all primary tumors, but for lymph node and distant metastases, they were only available with curative surgery or when there was doubt about the nature of the lesion (Table 1).
Published online before print 10.1148/radiol.14140090 Content code: Radiology 2014; 273:859–869 Abbreviations: AUC = area under the curve CI = confidence interval FDG = fluorine 18 fluorodeoxyglucose ICC = intraclass correlation coefficient LAVA = liver accelerated volume acquisition PROPELLER = periodically rotated overlapping parallel lines with enhanced reconstruction RECIST = Response Evaluation Criteria in Solid Tumors STIR = short inversion time inversion-recovery Author contributions: Guarantor of integrity of entire study, P.V.H.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, M.W.H., P.A., F.P.K., L.H., C.M.P., P.V.H.; clinical studies, M.W.H., P.A., F.P.K., L.H., C.M.P., I.A.B., M.P., G.D., P.V.H.; experimental studies, M.W.H., F.P.K., C.M.P.; statistical analysis, M.W.H., C.M.P.; and manuscript editing, M.W.H., F.P.K., L.H., C.M.P., I.A.B., G.K.v.S., P.V.H. Conflicts of interest are listed at the end of this article.
radiology.rsna.org n Radiology: Volume 273: Number 3—December 2014
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Table 1 Study Cohort and Standard of Reference No. of Lesions† Tumor Type Bronchial carcinoma Malignant melanoma Breast cancer Prostate cancer Esophageal cancer Colon cancer Thyroid cancer Hodgkin lymphoma Non-Hodgkin lymphoma Squamous cell cancer of the skin Gastric cancer Renal cell cancer Adrenocortical carcinoma Endometrial cancer Ovarian cancer Transitional cell cancer of ureter Kaposi sarcoma Malignant peripheral nerve sheath tumor Multiple myeloma Cancer of unknown primary Total
No. of Subjects*
Primary Tumors
Lymph Node Metastases
Distant Metastases
35/14/21 19/2/17 18/2/16 6/2/4 5/0/5 2/0/2 2/1/1 2/0/2 2/1/1 2/1/1 1/0/1 1/0/1 1/0/1 1/0/1 1/0/1 1/0/1 1/0/1 1/1/0 1/0/1 4/2/2 106/26/80
29/29/0 2/2/0 5/5/0 0 2/2/0 0 0 0 1/1/0 0 0 1/1/0 1/1/0 0 0 1/1/0 0 1/1/0 0 0 43/43/0
36/20/16 6/4/2 8/1/7 3/1/2 5/0/5 1/0/1 3/3/0 3/0/3 2/1/1 0 2/2/0 0 0 0 1/1/0 0 0 0 0 4/3/1 74/36/38
20/4/16 4/2/2 23/5/18 2/2/0 3/0/3 2/0/2 4/2/2 0 0 1/1/0 0 1/1/0 0 0 0 0 1/1/0 2/2/0 0 3/2/1 66/22/44
Note.—Locations of lymph node metastases and distant metastases are given in Appendix E4 (online). Nodal manifestations of lymphoma are included within lymph node metastases. * Data are presented as total number of subjects/number of subjects referred for staging/number of subjects referred for restaging. †
Data are presented as total number of lesions/number of lesions with histopathologic findings as the reference standard/ number of lesions followed up.
Imaging Techniques All patients were imaged with a trimodality PET/CT/MR system, which consists of a full-ring time-of-flight 64-section PET/CT scanner (Discovery PET/ CT 690 VCT; GE Healthcare, Waukesha, Wis) and a 3-T MR imager (Discovery MR 750w; GE Healthcare), situated in adjacent rooms. Both machines were linked by a shuttle system (transfer device), which obviates repositioning between examinations. (Details are given in Appendix E2 [online].) For examinations with FDG, patients fasted at least 4 hours prior to injection of a standardized dose of 350 MBq of FDG. The total uptake time was standardized to 60 minutes. After 40 minutes, patients were placed in supine position on the shuttle board (required time approximately 1 minute) and then
transferred into the MR machine for the remaining approximately 19 minutes of the uptake time. This examination time was chosen so that the data acquisition could also be readily implemented in a fully integrated PET/MR system. After completion of the MR examination (approximately 16 minutes), patients were shuttled to the PET/CT machine, where acquisition was started subsequently at 60 minutes after injection. The total PET/CT acquisition time was 12–16 minutes. For examinations with fluorine 18 (18F) fluoroethylcholine, the procedure was essentially the same; however, no fasting and no uptake period were required. The administered standardized dose was 200 MBq of 18F-fluoroethylcholine. Intravenous contrast material was not given for CT or MR imaging.
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
Huellner et al
PET/CT.—PET and nonenhanced CT data sets were acquired from vertex to midthighs. All PET examinations were performed during shallow breathing. CT data were acquired for PET attenuation correction and diagnostic purposes. MR imaging.—After placing the patient onto the shuttle board that was lying on the MR table, a radiofrequency coil (GEM 32-channel torso coil, with posterior and anterior array combined; GE Healthcare) was positioned, and the patient was moved into the MR gantry. The MR protocol consisted of three pulse sequences, chosen by means of preliminary testing on volunteers. The sequences had to fulfill the following criteria: (a) maintain the same body coverage as CT within the PET uptake time, (b) contain one sequence for MRbased attenuation correction, and (c) include one higher-resolution sequence for the lung and upper abdomen. Whole-body multisection imaging was performed with an axial T1-weighted three-dimensional dual-echo gradient-recalled-echo sequence (liver accelerated volume acquisition [LAVA]-Flex; GE Healthcare) and a coronally acquired short inversion time inversion-recovery (STIR) sequence without breath-hold technique. For lung imaging, an axial T2-weighted sequence with motion correction (periodically rotated overlapping parallel lines with enhanced reconstruction [PROPELLER]; GE Healthcare) was performed during free breathing. Details are given in Table 2.
Image Evaluation All PET/CT and PET/MR image data sets were anonymized and evaluated in random order by dually trained radiologists and nuclear medicine physicians with several years of experience in both MR and PET/CT (M.W.H. [reader 1], with 5 years and 2 years of experience, respectively; P.A. [reader 2], with 6 years and 5 years of experience, respectively; and P.V.H. [reader 3], with 7 years and 10 years of experience, respectively). Reader 3, who was not engaged in any further image analysis, reviewed the PET/CT scans of all patients with evidence of tumor and, in a preselection 861
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
process, identified all lesions that were positive at PET and visible at PET/CT. The largest measurable lesions were identified on PET/CT images on the basis of Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 target lesion selection criteria (Fig 1) (6). Details regarding the assessment
Huellner et al
algorithm are given in Appendix E3 (online). The primary tumor (if any), up to two lymph node metastases, and up to two distant metastases were identified. Subsequently, the entire data sets of PET/CT and PET/MR were evaluated completely for lesions relevant with RECIST and were assessed independently
Table 2 Acquisition Parameters for MR Imaging Parameter Repetition time (msec)/echo time (msec) Flip angle (degrees) Partial Fourier transform (%) Inversion time (msec) Parallel imaging acceleration factor Section thickness (mm) Field of view (cm) Acquisition matrix (pixels) Receiver bandwidth (kHz) Acquisition time per body section (sec) Body sections per patient Approximate total acquisition time (min) Coverage
Note.—NA = not applicable.
LAVA
STIR
PROPELLER
4.3/1.3 (opposed phase), 4.3/2.6 (in phase) 12 0.9 NA 2 4.0 50 288 3 224 142.86 18 4 3 Whole body
2000/42
9321/122
NA NA 160 2 6 50 384 3 224 100 123 3 8 Whole body
NA NA NA 3 4.5 40 288 3 288 62.5 NA 1 5 Chest, upper abdomen
by reader 1 and reader 2, both of whom were blinded to patient data. Both readers analyzed all of the predefined RECIST-based lesions on all images that showed the lesion in all three planes (three-dimensional data sets) or on images that showed the lesion in one plane (two-dimensional data sets). They determined the conspicuity of each lesion on a five-point scale (0, not visible; 1, poorly identified [circumferential delineation with less than 25% of lesion borders definable]; 2, moderately identified [25%–50% of borders definable]; 3, well identified [50%–75% of borders definable]; and 4, excellently identified [.75% of borders definable]). Furthermore, image-degrading artifacts were assessed on a three-point scale (0, no artifacts; 1, mild artifacts [not interfering with assessment]; and 2, severe artifacts [unsuitable for assessment]). The readers’ diagnostic confidence was determined by using a nominal scale to assess which modality was found to be more useful (PET/MR more useful, both PET/MR and PET/CT equally useful, or PET/CT more useful). As a next step, readers independently assessed all remaining (ie, not preselected) lesions that were positive
Figure 1
Figure 1: Flowchart shows the image analysis algorithm. The double-headed arrow links categories with identical lesions. 862
radiology.rsna.org n Radiology: Volume 273: Number 3—December 2014
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Huellner et al
Table 3 FDG PET Parameters of Primary Tumors, Lymph Node Metastases, and Distant Metastases, Including Subanalyses for Lung Tumors and Lung Metastases Primary Tumors FDG PET Parameters SUVmax* SUVmean(42%)* Total lesion glycolysis† PETvol (cm3)†
All (n = 43)
Lung Tumors (n = 29)
Lymph Node Metastases (n = 71)
13.1 6 10.1 7.8 6 6.1 38.1 (2.2–3200.1) 6.3 (0.8–300.0)
12.5 6 9.2 7.3 6 5.5 46.8 (2.2–3200.1) 8.2 (0.8–300.0)
8.8 6 3.9 5.3 6 2.5 9.7 (1.9–155.6) 2.0 (0.3–32.2)
Distant Metastases All (n = 64)
Lung Metastases (n = 17)
11.1 6 8.9 6.8 6 5.5 13.1 (0.3–394.7) 2.5 (0.2–63.5)
10.9 6 9.6 6.7 6 6.2 9.2 (0.3–179.1) 1.8 (0.4–41.3)
Note.—Three lymph node metastases and two distant metastases of patients imaged with 18F-fluoroethylcholine were not included. Nodal manifestations of lymphoma are included within lymph node metastases. PETvol = metabolic tumor volume, SUVmax = maximum standardized uptake value, SUVmean(42%) = mean standardized uptake value (the threshold of 42% means that all voxels of 42% or more of the SUVmax are included in the calculation of total lesion glycolysis and PETvol of the tumor volume). * Data are means 6 standard deviations. †
Data are median values, with ranges in parentheses.
at PET for further characterization (related or unrelated to the oncologic disease). Finally, all relevant incidental nononcologic findings on the different image sets were evaluated in consensus. Such findings (positive or negative at PET) were defined as being unrelated to the primary disease but could have been mistaken for neoplasia or a lesion, the treatment or knowledge of which could have an effect on the patient’s general condition or treatment.
Statistical Analysis Ordinal nondichotomous variables were expressed as median (with range), nominal nondichotomous variables were expressed as mode (percentage), and ratio variables were expressed as geometric mean 6 standard deviation. The agreement between both readers concerning lesion size and the agreement between CT and MR performed with different sequences was assessed by using t tests for related samples and intraclass correlation analysis. Agreement between the readers concerning conspicuity, image artifacts, and confidence was assessed by using k statistics. Agreement was defined as moderate (k = 0.41–0.60), substantial (k = 0.61–0.80), or almost perfect (k . 0.80) (7). Size and conspicuity of lesions and degree of artifacts with different modalities were compared by using a Wilcoxon signed rank test for matched pairs. A corresponding subanalysis was performed
for pulmonary lesions (no adjustment was made for multiple comparisons). As CT is the imaging standard of reference for lung lesions, the area under the receiver operating characteristic curve, fitted for lesion size in CT (continuous) and conspicuity with different MR sequences (categorical), was calculated by using a nonparametric model. The area under the curve (AUC) and the 95% confidence intervals (CIs) were calculated. The reader confidence with different modalities was compared by using a one-sample KolmogorovSmirnov test. Differences in numbers and types of incidental findings detected with both modalities were addressed with the Mann-Whitney U test and the Wilcoxon signed rank test, respectively. A P value less than .05 was considered to indicate a significant difference. For all statistical analyses, IBM SPSS Statistics software, version 19.0.1 (IBM, Armonk, NY), was used.
Results Thirty-one patients without evidence of tumor at preliminary analysis were excluded from further analysis. Of the remaining 75, 43 had a primary tumor, 43 had at least one lymph node metastasis (including nodal manifestations of lymphoma), and 38 had at least one distant metastasis. Primary tumors were lung cancer (n = 29), breast cancer (n = 5), malignant melanoma (n = 2), esophageal cancer (n
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
= 2), ureteral transitional cell cancer (n = 1), renal cell cancer (n = 1), adrenocortical carcinoma (n = 1), non-Hodgkin lymphoma (n = 1), and malignant peripheral nerve sheath tumor (n = 1). (Locations of lymph node and distant metastases are given in Appendix E4 [online].) Thus, a total of 183 RECIST-based lesions, consisting of 43 primary tumors, 74 lymph node metastases, and 66 distant metastases, were evaluated. The PROPELLER sequence that covered only the chest and parts of the upper abdomen did not depict 69 lesions. The STIR sequence did not depict six lesions, owing to technical issues. Detailed results of 18F-FDG PET are given in Table 3. In addition, the readers detected 64 oncologic lesions that were positive at PET beyond the RECIST-based lesions (27 nodal metastases and 37 distant metastases).
Interreader Agreement High interreader agreement was found for lesion conspicuity (CT, k = 0.90; STIR, k = 0.79; LAVA, k = 0.83; and PROPELLER, k = 0.84), image artifacts (CT, k = 0.87; STIR, k = 0.83; LAVA, k = 0.84; and PROPELLER, k = 0.84), and confidence (k = 0.87), and high intraclass correlation coefficients (ICCs) were found for lesion size (CT and all MR sequences, ICC = 0.99 [P , .001] each). Therefore, mean size measurements and mean scores for lesion conspicuity, image artifacts, and confidence 863
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Huellner et al
Table 4 Assessment of Lesions with CT, STIR, LAVA, and PROPELLER STIR Tumor Type and Parameter Primary tumors All No. of tumors Size (mm) ICC for size (MR vs CT) Conspicuity Artifacts Lung tumors No. of tumors Size (mm) Conspicuity Artifacts Lymph node metastases No. of metastases Size (mm) ICC for size (MR vs CT) Conspicuity Artifacts Distant metastases All No. of metastases Size (mm) ICC for size (MR vs CT) Conspicuity Artifacts Lung metastases No. of metastases Size (mm) Conspicuity Artifacts
LAVA
PROPELLER
CT
Value
P Value
Value
P Value
Value
P Value
43 37.4 6 21.6 NA 3.49 6 0.91 0.08 6 0.31
42 36.5 6 21.6 0.96 3.08 6 1.13 0.78 6 0.69
...
43 39.1 6 21.7 0.94 2.64 6 1.43 0.54 6 0.71
...
35 38.6 6 22.9 0.97 3.70 6 0.66 0.24 6 0.43
...
29 39.5 6 22.3 3.71 6 0.79 0.05 6 0.28
28 37.7 6 22.5 3.32 6 0.98 0.85 6 0.71
...
29 42.7 6 21.8 2.52 6 1.58 0.75 6 0.72
...
28 37.0 6 22.9 3.64 6 0.72 0.27 6 0.44
...
.254 .002 .010
74 11.8 6 4.9 NA 3.03 6 0.95 0.09 6 0.32
71 12.6 6 4.9 0.91 2.36 6 1.35 0.75 6 0.56
...
74 12.0 6 5.1 0.89 3.01 6 0.93 0.34 6 0.54
...
48 13.4 6 5.9 0.80 3.24 6 1.08 0.49 6 0.58
...
.989 ,.001 .986 .001
66 24.2 6 15.7 NA 2.88 6 1.31 0.16 6 0.45
64 24.5 6 14.7 0.97 3.01 6 1.14 0.39 6 0.57
...
66 25.6 6 15.9 0.95 2.78 6 1.31 0.23 6 0.41
...
31 20.6 6 11.9 0.93 3.84 6 0.54 0.15 6 0.35
...
17 22.1 6 14.3 3.56 6 0.71 0.06 6 0.24
17 21.3 6 15.1 2.88 6 1.23 0.66 6 0.60
...
17 22.5 6 16.4 2.12 6 1.40 0.61 6 0.56
...
16 22.2 6 14.9 4.00 6 0.00 0.03 6 0.13
...
.026 ,.001 .016 ,.001
.018 .014 ,.001
.131 ,.001 .001 ,.001
.145 ,.001 .527 .030
.172 .032 .005
.179 ,.001 .002 .011
.036 ,.001 .708 .400
.135 .005 .011
.096 ,.001 .436 .029
.071 .546 .001
.561 ,.001 .225 ,.001
.936 ,.001 ,.001 .667
.323 .026 .655
Note.—Data are means 6 standard deviations, unless indicated otherwise. Size is given as maximum lesion diameter, conspicuity as a five-point scale score (0–4), and artifacts as a three-point scale score (0–2). Nodal manifestations of lymphoma are included within lymph node metastases. P values refer to comparison with CT. NA = not applicable.
of the two readers were considered for further analysis (Table 4).
Primary Tumors The analysis showed high ICCs for the size of primary lung tumors on CT and MR images (Fig 2, Fig E1 [online]). Tumor size was significantly smaller on STIR images compared with CT images (P = .026; mean difference, 0.9 mm). Tumors were less conspicuous on STIR and LAVA images compared with CT images (P = .016 and P = .002, respectively) and were equally conspicuous on PROPELLER and CT images (Fig 3a). Images from all MR sequences had 864
significantly more artifacts than CT images, with STIR having the highest artifact score and PROPELLER having the lowest (Fig 3b).
Lymph Node Metastases ICCs for lymph node metastasis size on CT and MR images were high. No significant difference was found on CT and MR images. STIR images yielded a lower conspicuity of lymph node metastases compared with CT images (P = .001), while there was no difference between CT and LAVA or PROPELLER images (Fig 3c). MR images were degraded by artifacts to a higher extent than were CT images.
Distant Metastases High ICCs were found for the size of distant metastases on CT and MR images (Fig E2 [online]). The size of distant metastases was significantly larger on STIR images compared with CT images (P = .036; mean difference, 1.4 mm). No difference was found in lesion conspicuity on CT images versus STIR or LAVA images. However, there was a significantly higher distant metastasis conspicuity on PROPELLER images compared with CT images (conspicuity score for PROPELLER, 3.84; score for CT, 2.88; P , .001) (Fig 3d). Only STIR images showed more artifacts than CT images (P = .030), while there
radiology.rsna.org n Radiology: Volume 273: Number 3—December 2014
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Huellner et al
Figure 2
Figure 2: Images in a 65-year-old man with stage T3 lung cancer in the left upper lobe (arrow). (a) Axial CT image in lung window and (b) coregistered PET/CT image show a spiculated mass with pleural projections. Spiculae of the lesion are less well appreciated on the (c) axial LAVA (water only) image (repetition time [msec]/echo time [msec], 4.3/1.3; flip angle, 12°) and (d) coregistered PET/LAVA image, as well as on the (e) coronal STIR image (2000/42; inversion time, 160 msec) and (f) coregistered PET/ STIR image. The character of the lesion is nicely seen on the (g) axial PROPELLER image (9321/122) and (h) coregistered PET/PROPELLER image.
was no significant difference between CT images and LAVA or PROPELLER images (Fig 3e).
Lung Lesions The subanalysis of primary lung tumors yielded essentially the same results as those for all primary tumors, with a tendency toward higher artifact scores on MR images (Fig 4a). For lung metastases, however, the subanalysis yielded higher conspicuity for PROPELLER images than CT images (P = .026), while the conspicuity of STIR and LAVA images was significantly lower compared with CT images (P = .032 and P = .005, respectively) (Fig 4b). More artifacts were found on STIR and LAVA images
than on CT images (P = .005 and P = .011, respectively), while there was no significant difference between PROPELLER images and CT images. Lung lesions smaller than 25 mm (n = 23) were found to be less conspicuous than larger lesions on LAVA images (P , .001), but there was no such difference with STIR, PROPELLER, or CT images (Fig 4c). The receiver operating characteristic analysis demonstrated that the size of a lung lesion significantly influenced its conspicuity on LAVA images (AUC = 0.78; 95% CI: 0.64, 0.82; P = .002) and PROPELLER images (AUC = 0.98; 95% CI: 0.93, 1.00; P = .006) but not on STIR images (AUC = 0.70; 95% CI: 0.49, 0.91; P = .10) (Fig 4d).
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
Reader Confidence The readers were more confident with the diagnostic information derived from PET/MR images than that from PET/CT images in nine of 43 primary tumors (21%), with the reverse in two primary tumors (5%) and no difference in the remaining 32 cases (74%). Readers were more confident with PET/MR than with PET/CT in nine lymph node metastases (12%), more confident with PET/CT versus PET/MR in one lesion (1%), and equally confident in the remaining 64 lesions (86%). Regarding distant metastases, higher confidence was found with PET/MR than with PET/CT in 17 of 66 lesions (26%), higher confidence was 865
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Huellner et al
Figure 3
Figure 3: Graphs illustrate lesion conspicuity and image artifacts in primary tumors, lymph node metastases, and distant metastases. (a) Mean conspicuity of all primary tumors is depicted for CT images, STIR images (n = 42), LAVA images (n = 43), and PROPELLER images (n = 35). (b) Mean image artifacts are depicted for all primary tumors on CT and STIR images (n = 42), LAVA images (n = 43), and PROPELLER images (n = 35). (c) Mean conspicuity of local lymph node metastases is depicted for CT and STIR images (n = 71), LAVA images (n = 74), and PROPELLER images (n = 48). (d) Mean conspicuity of all distant metastases is depicted for CT and STIR images (n = 64), LAVA images (n = 66), and PROPELLER images (n = 31). (e) Mean image artifacts are depicted for all distant metastases on CT and STIR images (n = 64), LAVA images (n = 66), and PROPELLER images (n = 31).
found with PET/CT than with PET/MR in three lesions (4%), and equal confidence was found in the remaining 46 lesions (70%) (Fig E3 [online]). Reader confidence with PET/MR images was significantly higher than that with PET/ CT images in primary tumors (P = .02), lymph node metastases (P = .005), and distant metastases (P , .001).
Incidental Nononcologic Findings Seventy-two potentially relevant incidental nononcologic findings were detected, and significantly more were found with PET/CT than with PET/MR (total, 72; PET/CT, 66; and PET/MR, 55; P = 866
.039). PET/CT demonstrated 17 lesions more than PET/MR, and PET/MR demonstrated six lesions more than PET/CT. Significantly more nononcologic findings were detected with PET/CT than with PET/MR in the lung (total, 11; PET/CT, 11; and PET/MR, 0; P , .001). Other nononcologic findings consisted of lesions in the brain (total, three; PET/CT, three; and PET/MR, three), pleura (total, six; PET/CT, five; and PET/MR, six), heart and vessels (total, seven; PET/ CT, six; and PET/MR, six), liver and pancreas (total, 12; PET/CT, nine; and PET/MR, 11; P = .514), gastrointestinal tract (total, 11; PET/CT, 10; and PET/
MR, 11; P = .748), urogenital tract (total, six; PET/CT, six; and PET/MR, five), salivary and thyroid glands (total, six; PET/CT, six; and PET/MR, four), bone (total, four; PET/CT, four; and PET/MR, four), and soft tissues and breast (total, six; PET/CT, six; and PET/MR, five). All incidental nononcologic findings were negative at PET, except for two Warthin tumors and one foreign-body reaction to a vascular graft.
Discussion This prospective study in patients with oncologic diseases was conducted to
radiology.rsna.org n Radiology: Volume 273: Number 3—December 2014
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Huellner et al
Figure 4
Figure 4: Graphs illustrate lesion conspicuity and image artifacts in lung lesions. (a) Mean conspicuity of primary lung tumors is depicted for CT and STIR images (n = 28), LAVA images (n = 29), and PROPELLER images (n = 28). (b) Mean conspicuity of distant metastases to the lung is depicted for CT and STIR images (n = 17), LAVA images (n = 17), and PROPELLER images (n = 16). (c) Lung lesion conspicuity is depicted for CT images and images acquired with different MR sequences according to lung lesion size at CT. Graphs display size versus conspicuity of all lung lesions on CT and MR images, after averaging in bins of five lesions consecutive in size at CT. (d) Receiver operating characteristic curves are demonstrated for lung lesion size versus conspicuity with different MR sequences. Sufficient conspicuity was therefore defined as having a score of at least 3 (LAVA: AUC = 0.78; 95% CI: 0.64, 0.82 [P = .002]; PROPELLER: AUC = 0.98; 95% CI: 0.93, 1.00 [P = .006]; and STIR: AUC = 0.70; 95% CI: 0.49, 0.91 [P = .10]).
compare whole-body nonenhanced PET/MR and PET/CT by using similar acquisition times to identify MR pulse sequences that may be used most effectively in hybrid imaging with PET and to compare lesion detection of PET/MR with that of PET/CT.
We found that PET/MR with STIR and LAVA sequences is inferior to PET/ CT in terms of lesion conspicuity of primary tumors, including lung tumors, while PET/MR with the PROPELLER sequence yielded comparable conspicuity. The conspicuity of lymph node
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
metastases was generally lower than that of primary tumors. Wiesmüller and coworkers showed an equal performance of PET/MR and PET/CT in lesion detection in 46 patients with mainly metastatic lesions (8). No significant difference in detection rates at PET/ 867
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
MR versus PET/CT with a lesion-based and a patient-based approach was found by Drzezga and coworkers (2). Ohno et al found STIR to be as useful as PET/ CT in thoracic lymph nodes, which corresponds to our results (9). Another study showed an advantage of PET/CT over PET/MR by using a low-resolution, Dixon-based attenuation correction sequence in 12 of 38 analyzed small lymph node metastases (5). Appenzeller et al demonstrated a higher conspicuity of lymph node metastases (n = 57) at PET/CT than at PET/MR with LAVA by using a body coil rather than a torso coil, which suggests that torso coils are needed (1). PROPELLER provided better spatial resolution than LAVA and thinner section thickness than STIR. Since one of our goals was to evaluate a PET/MR protocol with an acquisition time that did not exceed that of standard PET/CT, coverage by the PROPELLER sequence was limited to the chest and upper abdomen. Although most of the distant metastases were encountered in these regions in our study, apart from the skeleton, our results for the PROPELLER sequence cannot be directly translated into whole-body PET/MR. Eiber et al found PET/CT to be superior to PET/MR in lung nodules that were positive at PET (5), while in another study, investigators found no significant difference in detecting nodules that were positive or negative at PET (4). Recent work by Chandarana et al showed that PET/CT was superior in small nodules and nodules negative at PET (10). While CT is widely recognized as the best modality for the texture evaluation of lung parenchyma, higherresolution sequences, such as PROPELLER, may limit the known shortcomings of MR to a certain extent (11). In larger pulmonary lesions, the higher intrinsic contrast of MR imaging may provide better conspicuity and delineation. This may in part explain our results, our cohort having slightly larger lung metastases (mean diameter, 22 mm) than those in other studies (smaller than 5 mm to 21 mm) (4,5,10). We could verify lung lesion size and conspicuity dependence on LAVA and PROPELLER images with receiver operating characteristic analysis. 868
Huellner et al
The trade-off of such a higher-resolution sequence is a longer imaging time. Image quality is a key factor. MR artifacts may interfere with the interpretation of findings in a trimodality system, but with an integrated PET/MR system, artifacts have an additional effect on PET/MR attenuation correction. In our study, we found significantly fewer artifacts with PET/CT than with PET/MR. This may be due to the location of the lymph node metastases in our cohort, as many were in areas prone to MR artifacts, such as the infrahyoid area of the neck and the chest. Kuhn et al found significantly more artifacts at PET/MR than at PET/CT in the infrahyoid area of the neck, resulting from respiration and swallowing (12). In our study, artifacts were comparable on PET/MR and PET/ CT images only in distant metastases (with LAVA and PROPELLER), including the subanalysis of lung metastases (PROPELLER only). The results for LAVA concerning primary tumors and distant metastases may be explained by the fact that most of the primary tumors were lung tumors and hence were prone to artifacts on LAVA images. In contrast, many distant metastases were located in body parts that showed no involuntary movement, such as bone and subcutaneous tissue. The overall diagnostic confidence with PET/MR was far higher than that with PET/CT. Similar results were found by Kuhn at el (12,13). This is likely due to the high soft-tissue contrast in MR compared with CT, which facilitates lesion characterization. Readers were less confident with bone lesions on PET/MR images, which is at variance with the known high confidence in high-spatial-resolution MR for neoplastic bone lesions. Reporting on incidental findings is an integral part of any imaging examination. Incidental findings may affect the course of disease, as they might be mistaken for malignancy or may require additional therapy. In total, PET/ CT demonstrated 17 findings more than PET/MR, potentially having an effect on patient management, but did not demonstrate six that were detected with PET/MR. This is partly in contrast
to the recent findings by Catalano et al, who reported that PET/MR has a higher clinical effect than PET/CT (14). However, they used a set of different MR protocols in their subjects, contrast material enhanced and nonenhanced, which were individually adapted to the patient’s disease. Unlike in our study, they acquired fully diagnostic MR images, generating an examination time of 66 minutes 6 12, whereas our PET/MR protocol was limited to the time range of a clinical PET/CT protocol. Our findings suggest that STIR PET/ MR is inferior to PET/CT, although STIR has been advocated for lung tumors in the pre-PET/MR era (9,15,16). LAVA PET/MR (ie, Dixon-type PET/MR) was only inferior to PET/CT in primary tumors, was less prone to artifacts than STIR, and is known to provide good anatomic localization (2,5). As the key questions asked in hybrid PET examinations concern lymph node and distant metastasis status, and since LAVA provides valuable information about tissue composition and seems to be a good sequence for MR-based attenuation correction in fully integrated systems, whole-body LAVA seems to be a valuable component of a PET/MR examination. Use of the PROPELLER sequence may provide an edge over PET/CT, even when used only in the chest and upper abdomen at whole-body PET/MR. Additional points, such as the rationale for the selection of MR pulse sequences, the conspicuity of small lesions, the issue with pulmonary nodules, and the significance of additional findings, are discussed in Appendix E5 (online). A limitation of our study is the inhomogeneous patient population regarding primary tumors and metastases, which obviates drawing conclusions about specific tumor entities. However, the goal of our study was to evaluate a number of MR sequences to design a valuable oncologic PET/MR protocol and to ensure that PET/MR and PET/ CT were comparable. A second limitation is that for ethical reasons, histologic findings were not available for all lymph nodes and distant metastases. Third, the analysis based on RECIST
radiology.rsna.org n Radiology: Volume 273: Number 3—December 2014
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
target lesion selection criteria might have introduced bias. However, this approach ensured that the most important lesions for oncologic staging and restaging were assessed. Fourth, the study was performed without the use of contrast agents. However, contrast agents are not used consistently in PET/CT anyway, and we wanted to see whether a PET/MR protocol could be devised that matches the performance of a nonenhanced PET/CT protocol. Fifth, we also did not use MR-based attenuation correction for PET/MR, as CT-based attenuation correction was available. This drawback affects mostly therapy monitoring, where reliable attenuation correction is needed for comparative quantification. Furthermore, it can be assumed that the PET/MR attenuation correction issues will be overcome. In summary, our preliminary evaluation indicates that PET/MR with LAVA and PROPELLER performs comparably to PET/CT in whole-body oncologic imaging and neoplastic lung disease. The basic protocol proposed here, performed simultaneously with a PET/ MR system, would take no longer than PET/CT. Drawbacks of PET/MR are increased numbers of artifacts, which do not interfere with image interpretation, and fewer incidental nonneoplastic lung findings than with PET/CT, the detection of which has limited value and is frequently inconsequential. Further studies will be needed to evaluate whether selective PET/MR protocols with sequences tailored to certain tumors and body regions yield additional advantages of PET/MR over PET/CT without making PET/MR an excessively lengthy and therefore expensive imaging modality.
I.A.B. disclosed no relevant relationships. M.P. disclosed no relevant relationships. G.D. Activities related to the present article: author is an employee of GE Healthcare. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. G.K.v.S. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: author received research grants and payment from GE Healthcare for lecturing; author received payment from E. Lilly and Icon Medical Imaging for consulting. Other relationships: disclosed no relevant relationships. P.V.H. Activities related to the present article: institution received a grant from GE Healthcare. Activities not related to the present article: institution received grants from Siemens Medical Solutions, Philips Medical, and Roche Pharma; author received payment from GE Healthcare for speaking. Other relationships: disclosed no relevant relationships.
Disclosures of Conflicts of Interest: M.W.H. Activities related to the present article: institution received a grant from GE Healthcare. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. P.A. disclosed no relevant relationships. F.P.K. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: author received payment for travel/accommodations/ meeting expenses from GE Healthcare Switzerland. Other relationships: disclosed no relevant relationships. L.H. disclosed no relevant relationships. C.M.P. disclosed no relevant relationships.
6. Eisenhauer EA, Therasse P, Bogaerts J, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45(2):228–247.
References 1. Appenzeller P, Mader C, Huellner MW, et al. PET/CT versus body coil PET/MRI: how low can you go? Insights Imaging 2013;4(4): 481–490. 2. Drzezga A, Souvatzoglou M, Eiber M, et al. First clinical experience with integrated whole-body PET/MR: comparison to PET/ CT in patients with oncologic diagnoses. J Nucl Med 2012;53(6):845–855. 3. Yi CA, Shin KM, Lee KS, et al. Non-small cell lung cancer staging: efficacy comparison of integrated PET/CT versus 3.0-T wholebody MR imaging. Radiology 2008;248(2): 632–642. 4. Stolzmann P, Veit-Haibach P, Chuck N, et al. Detection rate, location, and size of pulmonary nodules in trimodality PET/CT-MR: comparison of low-dose CT and Dixon-based MR imaging. Invest Radiol 2013;48(5): 241–246. 5. Eiber M, Martinez-Möller A, Souvatzoglou M, et al. Value of a Dixon-based MR/PET attenuation correction sequence for the localization and evaluation of PET-positive lesions. Eur J Nucl Med Mol Imaging 2011;38(9): 1691–1701.
7. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33(1):159–174. 8. Wiesmüller M, Quick HH, Navalpakkam B, et al. Comparison of lesion detection and quantitation of tracer uptake between PET from a simultaneously acquiring wholebody PET/MR hybrid scanner and PET
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
Huellner et al
from PET/CT. Eur J Nucl Med Mol Imaging 2013;40(1):12–21. 9. Ohno Y, Koyama H, Nogami M, et al. STIR turbo SE MR imaging vs. coregistered FDGPET/CT: quantitative and qualitative assessment of N-stage in non-small-cell lung cancer patients. J Magn Reson Imaging 2007; 26(4):1071–1080. 10. Chandarana H, Heacock L, Rakheja R, et al. Pulmonary nodules in patients with primary malignancy: comparison of hybrid PET/MR and PET/CT imaging. Radiology 2013;268(3):874–881. 11. Both M, Schultze J, Reuter M, et al. Fast T1- and T2-weighted pulmonary MR-imaging in patients with bronchial carcinoma. Eur J Radiol 2005;53(3):478–488. 12. Kuhn FP, Hüllner M, Mader CE, et al. Contrast-enhanced PET/MR imaging versus contrast-enhanced PET/CT in head and neck cancer: how much MR information is needed? J Nucl Med 2014;55(4):551–558. 13. Kuhn FP, Crook DW, Mader CE, Appen zeller P, von Schulthess GK, Schmid DT. Discrimination and anatomical mapping of PET-positive lesions: comparison of CT attenuation-corrected PET images with coregistered MR and CT images in the abdomen. Eur J Nucl Med Mol Imaging 2013;40(1): 44–51. 14. Catalano OA, Rosen BR, Sahani DV, et al. Clinical impact of PET/MR imaging in patients with cancer undergoing same-day PET/CT: initial experience in 134 patients— a hypothesis-generating exploratory study. Radiology 2013;269(3):857–869. 15. Ohno Y, Koyama H, Onishi Y, et al. Nonsmall cell lung cancer: whole-body MR examination for M-stage assessment—utility for whole-body diffusion-weighted imaging compared with integrated FDG PET/CT. Radiology 2008;248(2):643–654. 16. Morikawa M, Demura Y, Ishizaki T, et al. The effectiveness of 18F-FDG PET/CT combined with STIR MRI for diagnosing nodal involvement in the thorax. J Nucl Med 2009; 50(1):81–87. 17. Ma J. Breath-hold water and fat imaging using a dual-echo two-point Dixon technique with an efficient and robust phase-correction algorithm. Magn Reson Med 2004; 52(2):415–419. 18. Pipe JG. Motion correction with PROPELLER MRI: application to head motion and free-breathing cardiac imaging. Magn Reson Med 1999;42(5):963–969.
869
Purpose:
To assess the diagnostic performance of whole-body non– contrast material–enhanced positron emission tomography (PET)/magnetic resonance (MR) imaging and PET/computed tomography (CT) for staging and restaging of cancers and provide guidance for modality and sequence selection.
Materials and Methods:
This study was approved by the institutional review board and national government authorities. One hundred six consecutive patients (median age, 68 years; 46 female and 60 male patients) referred for staging or restaging of oncologic malignancies underwent whole-body imaging with a sequential trimodality PET/CT/MR system. The MR protocol included short inversion time inversion-recovery (STIR), Dixon-type liver accelerated volume acquisition (LAVA; GE Healthcare, Waukesha, Wis), and respiratory-gated periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER; GE Healthcare) sequences. Primary tumors (n = 43), local lymph node metastases (n = 74), and distant metastases (n = 66) were evaluated for conspicuity (scored 0–4), artifacts (scored 0–2), and reader confidence on PET/CT and PET/MR images. Subanalysis for lung lesions (n = 46) was also performed. Relevant incidental findings with both modalities were compared. Interreader agreement was analyzed with intraclass correlation coefficients and k statistics. Lesion conspicuity, image artifacts, and incidental findings were analyzed with nonparametric tests.
Results:
Primary tumors were less conspicuous on STIR (3.08, P = .016) and LAVA (2.64, P = .002) images than on CT images (3.49), while findings with the PROPELLER sequence (3.70, P = .436) were comparable to those at CT. In distant metastases, the PROPELLER sequence (3.84) yielded better results than CT (2.88, P , .001). Subanalysis for lung lesions yielded similar results (primary lung tumors: CT, 3.71; STIR, 3.32 [P = .014]; LAVA, 2.52 [P = .002]; PROPELLER, 3.64 [P = .546]). Readers classified lesions more confidently with PET/ MR than PET/CT. However, PET/CT showed more incidental findings than PET/MR (P = .039), especially in the lung (P , .001). MR images had more artifacts than CT images.
Conclusion:
PET/MR performs comparably to PET/CT in whole-body oncology and neoplastic lung disease, with the use of appropriate sequences. Further studies are needed to define regionalized PET/MR protocols with sequences tailored to specific tumor entities.
1
From the Department of Medical Radiology, Divisions of Nuclear Medicine (M.W.H., P.A., F.P.K., L.H., C.M.P., I.A.B., M.P., G.D., G.K.v.S., P.V.H.), Diagnostic and Interventional Radiology (P.A., F.P.K., P.V.H.), and Neuroradiology (M.W.H., F.P.K.), University Hospital Zurich, Rämistrasse 100, 8091 Zurich, Switzerland; and GE Healthcare, Waukesha, Wis (G.D.). Received January 23, 2014; revision requested March 17; revision received April 22; accepted June 10; final version accepted June 18. Address correspondence to M.W.H. (e-mail: [email protected]). RSNA, 2014
q
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
q RSNA, 2014
Online supplemental material is available for this article. 859
Medicine
Martin W. Huellner, MD Philippe Appenzeller, MD Félix P. Kuhn, MD Lars Husmann, MD Carsten M. Pietsch, MD Irene A. Burger, MD Miguel Porto, MRT Gaspar Delso, PhD Gustav K. von Schulthess, MD, PhD Patrick Veit-Haibach, MD
Original Research n Nuclear
Whole-Body Nonenhanced PET/ MR versus PET/CT in the Staging and Restaging of Cancers: Preliminary Observations1
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
T
o consider positron emission tomography (PET)/magnetic resonance (MR) imaging a clinical alternative to PET/computed tomography (CT), two major goals have to be achieved. First, efficient imaging protocols have to be developed, preferably not exceeding the examination time needed for PET/ CT. Second, the sensitivity and specificity of PET/MR and PET/CT should be comparable. Only then can PET/MR be considered a clinically viable alternative to PET/CT. This implies that complete MR data acquisition is performed during PET acquisition, which is approximately 10–20 minutes, or 2–4 minutes per bed position. Therefore, the key in clinical whole-body PET/MR must be the determination of MR protocols and the selection of adequate MR pulse sequences to provide information that is complementary and/or confirmatory to that of PET, rather than redundant. In contrast to CT, where protocol issues are essentially binary (low dose or high dose; contrast material enhanced or nonenhanced), in MR imaging, multiple pulse sequences can be chosen. Clinical studies are needed to compare PET/ CT with PET/MR to identify MR pulse sequences that are most complementary
Advances in Knowledge nn Our preliminary study suggests that PET/MR performs comparably to PET/CT in whole-body oncology (primary tumors, P = .436) and neoplastic lung disease (primary lung tumors, P = .546). nn PET/MR with a combination of a fast whole-body Dixon-type sequence and a regionalized (lung and upper abdomen) highresolution, respiratory-gated T2-weighted sequence might provide a slight advantage over PET/ CT by yielding a higher conspicuity of distant metastases. nn PET/MR exhibits more nonproblematic image artifacts (primary tumors, P = .029), while PET/CT leads to more incidental nononcologic findings (P = .039), especially in the lung (P , .001). 860
Huellner et al
to the PET examination and ideally not more time consuming than PET. Appenzeller et al demonstrated that MR/PET imaging with a body coil is inferior to PET/CT in oncology patients (1). Drzezga and coworkers reported no difference in lesion detection between PET/CT and PET/MR in 32 patients (2). However, assessment of lesion conspicuity and image-degrading artifacts was not part of their study. Whole-body MR was reported to have an accuracy similar to that of fluorine 18 fluorodeoxyglucose (FDG) PET/CT for non–small cell lung cancer T and N staging, with MR being superior in the detection of brain and liver metastases, while PET/ CT performed better in lymph node and soft-tissue metastases (3). Apparently, three-dimensional Dixon-based gradient-recalled-echo sequences, often used for MR-based attenuation correction, perform comparably to low-doseCT. A study performed with a sequential trimodality system showed that such a three-dimensional pulse sequence may be sufficient for lung imaging with PET/ MR (4). A similar breath-hold sequence was also found to be comparable to lowdose CT for the localization of oncologic lesions that were positive at PET in PET/ CT and coregistered MR (5). The performance of different nonenhanced MR pulse sequences versus nonenhanced CT with regard to lesion conspicuity and image-degrading artifacts has so far not been studied in a larger cohort. The purpose of this prospective study was to assess the diagnostic performance of whole-body nonenhanced PET/MR and PET/CT for staging and restaging of cancers to provide guidance for modality and sequence selection.
Materials and Methods This study was approved by the institutional review board and by national government authorities. All subjects
Implication for Patient Care nn In oncology patients, cancers can be effectively staged and restaged by using PET/MR instead of PET/ CT.
provided signed informed consent prior to the examinations. Financial support was provided by an institutional grant from GE Healthcare. Only non-GE employees had control of inclusion of data and information that might present a conflict of interest for authors who are employees of GE Healthcare. In this prospective study, 106 consecutive patients (46 female and 60 male patients; median age, 68 years; range, 16–87 years) referred for staging (n = 26) or restaging (n = 80) of different oncologic malignancies between April 2012 and October 2012 underwent imaging with a PET/CT/MR system (detailed information is provided in Appendix E1 [online]).
Reference Standard The standard of reference consisted of histologic findings, if available, and clinical and imaging follow-up (mean, 335 days; range, 290–512 days). Histologic findings were available for all primary tumors, but for lymph node and distant metastases, they were only available with curative surgery or when there was doubt about the nature of the lesion (Table 1).
Published online before print 10.1148/radiol.14140090 Content code: Radiology 2014; 273:859–869 Abbreviations: AUC = area under the curve CI = confidence interval FDG = fluorine 18 fluorodeoxyglucose ICC = intraclass correlation coefficient LAVA = liver accelerated volume acquisition PROPELLER = periodically rotated overlapping parallel lines with enhanced reconstruction RECIST = Response Evaluation Criteria in Solid Tumors STIR = short inversion time inversion-recovery Author contributions: Guarantor of integrity of entire study, P.V.H.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, M.W.H., P.A., F.P.K., L.H., C.M.P., P.V.H.; clinical studies, M.W.H., P.A., F.P.K., L.H., C.M.P., I.A.B., M.P., G.D., P.V.H.; experimental studies, M.W.H., F.P.K., C.M.P.; statistical analysis, M.W.H., C.M.P.; and manuscript editing, M.W.H., F.P.K., L.H., C.M.P., I.A.B., G.K.v.S., P.V.H. Conflicts of interest are listed at the end of this article.
radiology.rsna.org n Radiology: Volume 273: Number 3—December 2014
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Table 1 Study Cohort and Standard of Reference No. of Lesions† Tumor Type Bronchial carcinoma Malignant melanoma Breast cancer Prostate cancer Esophageal cancer Colon cancer Thyroid cancer Hodgkin lymphoma Non-Hodgkin lymphoma Squamous cell cancer of the skin Gastric cancer Renal cell cancer Adrenocortical carcinoma Endometrial cancer Ovarian cancer Transitional cell cancer of ureter Kaposi sarcoma Malignant peripheral nerve sheath tumor Multiple myeloma Cancer of unknown primary Total
No. of Subjects*
Primary Tumors
Lymph Node Metastases
Distant Metastases
35/14/21 19/2/17 18/2/16 6/2/4 5/0/5 2/0/2 2/1/1 2/0/2 2/1/1 2/1/1 1/0/1 1/0/1 1/0/1 1/0/1 1/0/1 1/0/1 1/0/1 1/1/0 1/0/1 4/2/2 106/26/80
29/29/0 2/2/0 5/5/0 0 2/2/0 0 0 0 1/1/0 0 0 1/1/0 1/1/0 0 0 1/1/0 0 1/1/0 0 0 43/43/0
36/20/16 6/4/2 8/1/7 3/1/2 5/0/5 1/0/1 3/3/0 3/0/3 2/1/1 0 2/2/0 0 0 0 1/1/0 0 0 0 0 4/3/1 74/36/38
20/4/16 4/2/2 23/5/18 2/2/0 3/0/3 2/0/2 4/2/2 0 0 1/1/0 0 1/1/0 0 0 0 0 1/1/0 2/2/0 0 3/2/1 66/22/44
Note.—Locations of lymph node metastases and distant metastases are given in Appendix E4 (online). Nodal manifestations of lymphoma are included within lymph node metastases. * Data are presented as total number of subjects/number of subjects referred for staging/number of subjects referred for restaging. †
Data are presented as total number of lesions/number of lesions with histopathologic findings as the reference standard/ number of lesions followed up.
Imaging Techniques All patients were imaged with a trimodality PET/CT/MR system, which consists of a full-ring time-of-flight 64-section PET/CT scanner (Discovery PET/ CT 690 VCT; GE Healthcare, Waukesha, Wis) and a 3-T MR imager (Discovery MR 750w; GE Healthcare), situated in adjacent rooms. Both machines were linked by a shuttle system (transfer device), which obviates repositioning between examinations. (Details are given in Appendix E2 [online].) For examinations with FDG, patients fasted at least 4 hours prior to injection of a standardized dose of 350 MBq of FDG. The total uptake time was standardized to 60 minutes. After 40 minutes, patients were placed in supine position on the shuttle board (required time approximately 1 minute) and then
transferred into the MR machine for the remaining approximately 19 minutes of the uptake time. This examination time was chosen so that the data acquisition could also be readily implemented in a fully integrated PET/MR system. After completion of the MR examination (approximately 16 minutes), patients were shuttled to the PET/CT machine, where acquisition was started subsequently at 60 minutes after injection. The total PET/CT acquisition time was 12–16 minutes. For examinations with fluorine 18 (18F) fluoroethylcholine, the procedure was essentially the same; however, no fasting and no uptake period were required. The administered standardized dose was 200 MBq of 18F-fluoroethylcholine. Intravenous contrast material was not given for CT or MR imaging.
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
Huellner et al
PET/CT.—PET and nonenhanced CT data sets were acquired from vertex to midthighs. All PET examinations were performed during shallow breathing. CT data were acquired for PET attenuation correction and diagnostic purposes. MR imaging.—After placing the patient onto the shuttle board that was lying on the MR table, a radiofrequency coil (GEM 32-channel torso coil, with posterior and anterior array combined; GE Healthcare) was positioned, and the patient was moved into the MR gantry. The MR protocol consisted of three pulse sequences, chosen by means of preliminary testing on volunteers. The sequences had to fulfill the following criteria: (a) maintain the same body coverage as CT within the PET uptake time, (b) contain one sequence for MRbased attenuation correction, and (c) include one higher-resolution sequence for the lung and upper abdomen. Whole-body multisection imaging was performed with an axial T1-weighted three-dimensional dual-echo gradient-recalled-echo sequence (liver accelerated volume acquisition [LAVA]-Flex; GE Healthcare) and a coronally acquired short inversion time inversion-recovery (STIR) sequence without breath-hold technique. For lung imaging, an axial T2-weighted sequence with motion correction (periodically rotated overlapping parallel lines with enhanced reconstruction [PROPELLER]; GE Healthcare) was performed during free breathing. Details are given in Table 2.
Image Evaluation All PET/CT and PET/MR image data sets were anonymized and evaluated in random order by dually trained radiologists and nuclear medicine physicians with several years of experience in both MR and PET/CT (M.W.H. [reader 1], with 5 years and 2 years of experience, respectively; P.A. [reader 2], with 6 years and 5 years of experience, respectively; and P.V.H. [reader 3], with 7 years and 10 years of experience, respectively). Reader 3, who was not engaged in any further image analysis, reviewed the PET/CT scans of all patients with evidence of tumor and, in a preselection 861
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
process, identified all lesions that were positive at PET and visible at PET/CT. The largest measurable lesions were identified on PET/CT images on the basis of Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 target lesion selection criteria (Fig 1) (6). Details regarding the assessment
Huellner et al
algorithm are given in Appendix E3 (online). The primary tumor (if any), up to two lymph node metastases, and up to two distant metastases were identified. Subsequently, the entire data sets of PET/CT and PET/MR were evaluated completely for lesions relevant with RECIST and were assessed independently
Table 2 Acquisition Parameters for MR Imaging Parameter Repetition time (msec)/echo time (msec) Flip angle (degrees) Partial Fourier transform (%) Inversion time (msec) Parallel imaging acceleration factor Section thickness (mm) Field of view (cm) Acquisition matrix (pixels) Receiver bandwidth (kHz) Acquisition time per body section (sec) Body sections per patient Approximate total acquisition time (min) Coverage
Note.—NA = not applicable.
LAVA
STIR
PROPELLER
4.3/1.3 (opposed phase), 4.3/2.6 (in phase) 12 0.9 NA 2 4.0 50 288 3 224 142.86 18 4 3 Whole body
2000/42
9321/122
NA NA 160 2 6 50 384 3 224 100 123 3 8 Whole body
NA NA NA 3 4.5 40 288 3 288 62.5 NA 1 5 Chest, upper abdomen
by reader 1 and reader 2, both of whom were blinded to patient data. Both readers analyzed all of the predefined RECIST-based lesions on all images that showed the lesion in all three planes (three-dimensional data sets) or on images that showed the lesion in one plane (two-dimensional data sets). They determined the conspicuity of each lesion on a five-point scale (0, not visible; 1, poorly identified [circumferential delineation with less than 25% of lesion borders definable]; 2, moderately identified [25%–50% of borders definable]; 3, well identified [50%–75% of borders definable]; and 4, excellently identified [.75% of borders definable]). Furthermore, image-degrading artifacts were assessed on a three-point scale (0, no artifacts; 1, mild artifacts [not interfering with assessment]; and 2, severe artifacts [unsuitable for assessment]). The readers’ diagnostic confidence was determined by using a nominal scale to assess which modality was found to be more useful (PET/MR more useful, both PET/MR and PET/CT equally useful, or PET/CT more useful). As a next step, readers independently assessed all remaining (ie, not preselected) lesions that were positive
Figure 1
Figure 1: Flowchart shows the image analysis algorithm. The double-headed arrow links categories with identical lesions. 862
radiology.rsna.org n Radiology: Volume 273: Number 3—December 2014
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Huellner et al
Table 3 FDG PET Parameters of Primary Tumors, Lymph Node Metastases, and Distant Metastases, Including Subanalyses for Lung Tumors and Lung Metastases Primary Tumors FDG PET Parameters SUVmax* SUVmean(42%)* Total lesion glycolysis† PETvol (cm3)†
All (n = 43)
Lung Tumors (n = 29)
Lymph Node Metastases (n = 71)
13.1 6 10.1 7.8 6 6.1 38.1 (2.2–3200.1) 6.3 (0.8–300.0)
12.5 6 9.2 7.3 6 5.5 46.8 (2.2–3200.1) 8.2 (0.8–300.0)
8.8 6 3.9 5.3 6 2.5 9.7 (1.9–155.6) 2.0 (0.3–32.2)
Distant Metastases All (n = 64)
Lung Metastases (n = 17)
11.1 6 8.9 6.8 6 5.5 13.1 (0.3–394.7) 2.5 (0.2–63.5)
10.9 6 9.6 6.7 6 6.2 9.2 (0.3–179.1) 1.8 (0.4–41.3)
Note.—Three lymph node metastases and two distant metastases of patients imaged with 18F-fluoroethylcholine were not included. Nodal manifestations of lymphoma are included within lymph node metastases. PETvol = metabolic tumor volume, SUVmax = maximum standardized uptake value, SUVmean(42%) = mean standardized uptake value (the threshold of 42% means that all voxels of 42% or more of the SUVmax are included in the calculation of total lesion glycolysis and PETvol of the tumor volume). * Data are means 6 standard deviations. †
Data are median values, with ranges in parentheses.
at PET for further characterization (related or unrelated to the oncologic disease). Finally, all relevant incidental nononcologic findings on the different image sets were evaluated in consensus. Such findings (positive or negative at PET) were defined as being unrelated to the primary disease but could have been mistaken for neoplasia or a lesion, the treatment or knowledge of which could have an effect on the patient’s general condition or treatment.
Statistical Analysis Ordinal nondichotomous variables were expressed as median (with range), nominal nondichotomous variables were expressed as mode (percentage), and ratio variables were expressed as geometric mean 6 standard deviation. The agreement between both readers concerning lesion size and the agreement between CT and MR performed with different sequences was assessed by using t tests for related samples and intraclass correlation analysis. Agreement between the readers concerning conspicuity, image artifacts, and confidence was assessed by using k statistics. Agreement was defined as moderate (k = 0.41–0.60), substantial (k = 0.61–0.80), or almost perfect (k . 0.80) (7). Size and conspicuity of lesions and degree of artifacts with different modalities were compared by using a Wilcoxon signed rank test for matched pairs. A corresponding subanalysis was performed
for pulmonary lesions (no adjustment was made for multiple comparisons). As CT is the imaging standard of reference for lung lesions, the area under the receiver operating characteristic curve, fitted for lesion size in CT (continuous) and conspicuity with different MR sequences (categorical), was calculated by using a nonparametric model. The area under the curve (AUC) and the 95% confidence intervals (CIs) were calculated. The reader confidence with different modalities was compared by using a one-sample KolmogorovSmirnov test. Differences in numbers and types of incidental findings detected with both modalities were addressed with the Mann-Whitney U test and the Wilcoxon signed rank test, respectively. A P value less than .05 was considered to indicate a significant difference. For all statistical analyses, IBM SPSS Statistics software, version 19.0.1 (IBM, Armonk, NY), was used.
Results Thirty-one patients without evidence of tumor at preliminary analysis were excluded from further analysis. Of the remaining 75, 43 had a primary tumor, 43 had at least one lymph node metastasis (including nodal manifestations of lymphoma), and 38 had at least one distant metastasis. Primary tumors were lung cancer (n = 29), breast cancer (n = 5), malignant melanoma (n = 2), esophageal cancer (n
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
= 2), ureteral transitional cell cancer (n = 1), renal cell cancer (n = 1), adrenocortical carcinoma (n = 1), non-Hodgkin lymphoma (n = 1), and malignant peripheral nerve sheath tumor (n = 1). (Locations of lymph node and distant metastases are given in Appendix E4 [online].) Thus, a total of 183 RECIST-based lesions, consisting of 43 primary tumors, 74 lymph node metastases, and 66 distant metastases, were evaluated. The PROPELLER sequence that covered only the chest and parts of the upper abdomen did not depict 69 lesions. The STIR sequence did not depict six lesions, owing to technical issues. Detailed results of 18F-FDG PET are given in Table 3. In addition, the readers detected 64 oncologic lesions that were positive at PET beyond the RECIST-based lesions (27 nodal metastases and 37 distant metastases).
Interreader Agreement High interreader agreement was found for lesion conspicuity (CT, k = 0.90; STIR, k = 0.79; LAVA, k = 0.83; and PROPELLER, k = 0.84), image artifacts (CT, k = 0.87; STIR, k = 0.83; LAVA, k = 0.84; and PROPELLER, k = 0.84), and confidence (k = 0.87), and high intraclass correlation coefficients (ICCs) were found for lesion size (CT and all MR sequences, ICC = 0.99 [P , .001] each). Therefore, mean size measurements and mean scores for lesion conspicuity, image artifacts, and confidence 863
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Huellner et al
Table 4 Assessment of Lesions with CT, STIR, LAVA, and PROPELLER STIR Tumor Type and Parameter Primary tumors All No. of tumors Size (mm) ICC for size (MR vs CT) Conspicuity Artifacts Lung tumors No. of tumors Size (mm) Conspicuity Artifacts Lymph node metastases No. of metastases Size (mm) ICC for size (MR vs CT) Conspicuity Artifacts Distant metastases All No. of metastases Size (mm) ICC for size (MR vs CT) Conspicuity Artifacts Lung metastases No. of metastases Size (mm) Conspicuity Artifacts
LAVA
PROPELLER
CT
Value
P Value
Value
P Value
Value
P Value
43 37.4 6 21.6 NA 3.49 6 0.91 0.08 6 0.31
42 36.5 6 21.6 0.96 3.08 6 1.13 0.78 6 0.69
...
43 39.1 6 21.7 0.94 2.64 6 1.43 0.54 6 0.71
...
35 38.6 6 22.9 0.97 3.70 6 0.66 0.24 6 0.43
...
29 39.5 6 22.3 3.71 6 0.79 0.05 6 0.28
28 37.7 6 22.5 3.32 6 0.98 0.85 6 0.71
...
29 42.7 6 21.8 2.52 6 1.58 0.75 6 0.72
...
28 37.0 6 22.9 3.64 6 0.72 0.27 6 0.44
...
.254 .002 .010
74 11.8 6 4.9 NA 3.03 6 0.95 0.09 6 0.32
71 12.6 6 4.9 0.91 2.36 6 1.35 0.75 6 0.56
...
74 12.0 6 5.1 0.89 3.01 6 0.93 0.34 6 0.54
...
48 13.4 6 5.9 0.80 3.24 6 1.08 0.49 6 0.58
...
.989 ,.001 .986 .001
66 24.2 6 15.7 NA 2.88 6 1.31 0.16 6 0.45
64 24.5 6 14.7 0.97 3.01 6 1.14 0.39 6 0.57
...
66 25.6 6 15.9 0.95 2.78 6 1.31 0.23 6 0.41
...
31 20.6 6 11.9 0.93 3.84 6 0.54 0.15 6 0.35
...
17 22.1 6 14.3 3.56 6 0.71 0.06 6 0.24
17 21.3 6 15.1 2.88 6 1.23 0.66 6 0.60
...
17 22.5 6 16.4 2.12 6 1.40 0.61 6 0.56
...
16 22.2 6 14.9 4.00 6 0.00 0.03 6 0.13
...
.026 ,.001 .016 ,.001
.018 .014 ,.001
.131 ,.001 .001 ,.001
.145 ,.001 .527 .030
.172 .032 .005
.179 ,.001 .002 .011
.036 ,.001 .708 .400
.135 .005 .011
.096 ,.001 .436 .029
.071 .546 .001
.561 ,.001 .225 ,.001
.936 ,.001 ,.001 .667
.323 .026 .655
Note.—Data are means 6 standard deviations, unless indicated otherwise. Size is given as maximum lesion diameter, conspicuity as a five-point scale score (0–4), and artifacts as a three-point scale score (0–2). Nodal manifestations of lymphoma are included within lymph node metastases. P values refer to comparison with CT. NA = not applicable.
of the two readers were considered for further analysis (Table 4).
Primary Tumors The analysis showed high ICCs for the size of primary lung tumors on CT and MR images (Fig 2, Fig E1 [online]). Tumor size was significantly smaller on STIR images compared with CT images (P = .026; mean difference, 0.9 mm). Tumors were less conspicuous on STIR and LAVA images compared with CT images (P = .016 and P = .002, respectively) and were equally conspicuous on PROPELLER and CT images (Fig 3a). Images from all MR sequences had 864
significantly more artifacts than CT images, with STIR having the highest artifact score and PROPELLER having the lowest (Fig 3b).
Lymph Node Metastases ICCs for lymph node metastasis size on CT and MR images were high. No significant difference was found on CT and MR images. STIR images yielded a lower conspicuity of lymph node metastases compared with CT images (P = .001), while there was no difference between CT and LAVA or PROPELLER images (Fig 3c). MR images were degraded by artifacts to a higher extent than were CT images.
Distant Metastases High ICCs were found for the size of distant metastases on CT and MR images (Fig E2 [online]). The size of distant metastases was significantly larger on STIR images compared with CT images (P = .036; mean difference, 1.4 mm). No difference was found in lesion conspicuity on CT images versus STIR or LAVA images. However, there was a significantly higher distant metastasis conspicuity on PROPELLER images compared with CT images (conspicuity score for PROPELLER, 3.84; score for CT, 2.88; P , .001) (Fig 3d). Only STIR images showed more artifacts than CT images (P = .030), while there
radiology.rsna.org n Radiology: Volume 273: Number 3—December 2014
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Huellner et al
Figure 2
Figure 2: Images in a 65-year-old man with stage T3 lung cancer in the left upper lobe (arrow). (a) Axial CT image in lung window and (b) coregistered PET/CT image show a spiculated mass with pleural projections. Spiculae of the lesion are less well appreciated on the (c) axial LAVA (water only) image (repetition time [msec]/echo time [msec], 4.3/1.3; flip angle, 12°) and (d) coregistered PET/LAVA image, as well as on the (e) coronal STIR image (2000/42; inversion time, 160 msec) and (f) coregistered PET/ STIR image. The character of the lesion is nicely seen on the (g) axial PROPELLER image (9321/122) and (h) coregistered PET/PROPELLER image.
was no significant difference between CT images and LAVA or PROPELLER images (Fig 3e).
Lung Lesions The subanalysis of primary lung tumors yielded essentially the same results as those for all primary tumors, with a tendency toward higher artifact scores on MR images (Fig 4a). For lung metastases, however, the subanalysis yielded higher conspicuity for PROPELLER images than CT images (P = .026), while the conspicuity of STIR and LAVA images was significantly lower compared with CT images (P = .032 and P = .005, respectively) (Fig 4b). More artifacts were found on STIR and LAVA images
than on CT images (P = .005 and P = .011, respectively), while there was no significant difference between PROPELLER images and CT images. Lung lesions smaller than 25 mm (n = 23) were found to be less conspicuous than larger lesions on LAVA images (P , .001), but there was no such difference with STIR, PROPELLER, or CT images (Fig 4c). The receiver operating characteristic analysis demonstrated that the size of a lung lesion significantly influenced its conspicuity on LAVA images (AUC = 0.78; 95% CI: 0.64, 0.82; P = .002) and PROPELLER images (AUC = 0.98; 95% CI: 0.93, 1.00; P = .006) but not on STIR images (AUC = 0.70; 95% CI: 0.49, 0.91; P = .10) (Fig 4d).
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
Reader Confidence The readers were more confident with the diagnostic information derived from PET/MR images than that from PET/CT images in nine of 43 primary tumors (21%), with the reverse in two primary tumors (5%) and no difference in the remaining 32 cases (74%). Readers were more confident with PET/MR than with PET/CT in nine lymph node metastases (12%), more confident with PET/CT versus PET/MR in one lesion (1%), and equally confident in the remaining 64 lesions (86%). Regarding distant metastases, higher confidence was found with PET/MR than with PET/CT in 17 of 66 lesions (26%), higher confidence was 865
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Huellner et al
Figure 3
Figure 3: Graphs illustrate lesion conspicuity and image artifacts in primary tumors, lymph node metastases, and distant metastases. (a) Mean conspicuity of all primary tumors is depicted for CT images, STIR images (n = 42), LAVA images (n = 43), and PROPELLER images (n = 35). (b) Mean image artifacts are depicted for all primary tumors on CT and STIR images (n = 42), LAVA images (n = 43), and PROPELLER images (n = 35). (c) Mean conspicuity of local lymph node metastases is depicted for CT and STIR images (n = 71), LAVA images (n = 74), and PROPELLER images (n = 48). (d) Mean conspicuity of all distant metastases is depicted for CT and STIR images (n = 64), LAVA images (n = 66), and PROPELLER images (n = 31). (e) Mean image artifacts are depicted for all distant metastases on CT and STIR images (n = 64), LAVA images (n = 66), and PROPELLER images (n = 31).
found with PET/CT than with PET/MR in three lesions (4%), and equal confidence was found in the remaining 46 lesions (70%) (Fig E3 [online]). Reader confidence with PET/MR images was significantly higher than that with PET/ CT images in primary tumors (P = .02), lymph node metastases (P = .005), and distant metastases (P , .001).
Incidental Nononcologic Findings Seventy-two potentially relevant incidental nononcologic findings were detected, and significantly more were found with PET/CT than with PET/MR (total, 72; PET/CT, 66; and PET/MR, 55; P = 866
.039). PET/CT demonstrated 17 lesions more than PET/MR, and PET/MR demonstrated six lesions more than PET/CT. Significantly more nononcologic findings were detected with PET/CT than with PET/MR in the lung (total, 11; PET/CT, 11; and PET/MR, 0; P , .001). Other nononcologic findings consisted of lesions in the brain (total, three; PET/CT, three; and PET/MR, three), pleura (total, six; PET/CT, five; and PET/MR, six), heart and vessels (total, seven; PET/ CT, six; and PET/MR, six), liver and pancreas (total, 12; PET/CT, nine; and PET/MR, 11; P = .514), gastrointestinal tract (total, 11; PET/CT, 10; and PET/
MR, 11; P = .748), urogenital tract (total, six; PET/CT, six; and PET/MR, five), salivary and thyroid glands (total, six; PET/CT, six; and PET/MR, four), bone (total, four; PET/CT, four; and PET/MR, four), and soft tissues and breast (total, six; PET/CT, six; and PET/MR, five). All incidental nononcologic findings were negative at PET, except for two Warthin tumors and one foreign-body reaction to a vascular graft.
Discussion This prospective study in patients with oncologic diseases was conducted to
radiology.rsna.org n Radiology: Volume 273: Number 3—December 2014
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
Huellner et al
Figure 4
Figure 4: Graphs illustrate lesion conspicuity and image artifacts in lung lesions. (a) Mean conspicuity of primary lung tumors is depicted for CT and STIR images (n = 28), LAVA images (n = 29), and PROPELLER images (n = 28). (b) Mean conspicuity of distant metastases to the lung is depicted for CT and STIR images (n = 17), LAVA images (n = 17), and PROPELLER images (n = 16). (c) Lung lesion conspicuity is depicted for CT images and images acquired with different MR sequences according to lung lesion size at CT. Graphs display size versus conspicuity of all lung lesions on CT and MR images, after averaging in bins of five lesions consecutive in size at CT. (d) Receiver operating characteristic curves are demonstrated for lung lesion size versus conspicuity with different MR sequences. Sufficient conspicuity was therefore defined as having a score of at least 3 (LAVA: AUC = 0.78; 95% CI: 0.64, 0.82 [P = .002]; PROPELLER: AUC = 0.98; 95% CI: 0.93, 1.00 [P = .006]; and STIR: AUC = 0.70; 95% CI: 0.49, 0.91 [P = .10]).
compare whole-body nonenhanced PET/MR and PET/CT by using similar acquisition times to identify MR pulse sequences that may be used most effectively in hybrid imaging with PET and to compare lesion detection of PET/MR with that of PET/CT.
We found that PET/MR with STIR and LAVA sequences is inferior to PET/ CT in terms of lesion conspicuity of primary tumors, including lung tumors, while PET/MR with the PROPELLER sequence yielded comparable conspicuity. The conspicuity of lymph node
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
metastases was generally lower than that of primary tumors. Wiesmüller and coworkers showed an equal performance of PET/MR and PET/CT in lesion detection in 46 patients with mainly metastatic lesions (8). No significant difference in detection rates at PET/ 867
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
MR versus PET/CT with a lesion-based and a patient-based approach was found by Drzezga and coworkers (2). Ohno et al found STIR to be as useful as PET/ CT in thoracic lymph nodes, which corresponds to our results (9). Another study showed an advantage of PET/CT over PET/MR by using a low-resolution, Dixon-based attenuation correction sequence in 12 of 38 analyzed small lymph node metastases (5). Appenzeller et al demonstrated a higher conspicuity of lymph node metastases (n = 57) at PET/CT than at PET/MR with LAVA by using a body coil rather than a torso coil, which suggests that torso coils are needed (1). PROPELLER provided better spatial resolution than LAVA and thinner section thickness than STIR. Since one of our goals was to evaluate a PET/MR protocol with an acquisition time that did not exceed that of standard PET/CT, coverage by the PROPELLER sequence was limited to the chest and upper abdomen. Although most of the distant metastases were encountered in these regions in our study, apart from the skeleton, our results for the PROPELLER sequence cannot be directly translated into whole-body PET/MR. Eiber et al found PET/CT to be superior to PET/MR in lung nodules that were positive at PET (5), while in another study, investigators found no significant difference in detecting nodules that were positive or negative at PET (4). Recent work by Chandarana et al showed that PET/CT was superior in small nodules and nodules negative at PET (10). While CT is widely recognized as the best modality for the texture evaluation of lung parenchyma, higherresolution sequences, such as PROPELLER, may limit the known shortcomings of MR to a certain extent (11). In larger pulmonary lesions, the higher intrinsic contrast of MR imaging may provide better conspicuity and delineation. This may in part explain our results, our cohort having slightly larger lung metastases (mean diameter, 22 mm) than those in other studies (smaller than 5 mm to 21 mm) (4,5,10). We could verify lung lesion size and conspicuity dependence on LAVA and PROPELLER images with receiver operating characteristic analysis. 868
Huellner et al
The trade-off of such a higher-resolution sequence is a longer imaging time. Image quality is a key factor. MR artifacts may interfere with the interpretation of findings in a trimodality system, but with an integrated PET/MR system, artifacts have an additional effect on PET/MR attenuation correction. In our study, we found significantly fewer artifacts with PET/CT than with PET/MR. This may be due to the location of the lymph node metastases in our cohort, as many were in areas prone to MR artifacts, such as the infrahyoid area of the neck and the chest. Kuhn et al found significantly more artifacts at PET/MR than at PET/CT in the infrahyoid area of the neck, resulting from respiration and swallowing (12). In our study, artifacts were comparable on PET/MR and PET/ CT images only in distant metastases (with LAVA and PROPELLER), including the subanalysis of lung metastases (PROPELLER only). The results for LAVA concerning primary tumors and distant metastases may be explained by the fact that most of the primary tumors were lung tumors and hence were prone to artifacts on LAVA images. In contrast, many distant metastases were located in body parts that showed no involuntary movement, such as bone and subcutaneous tissue. The overall diagnostic confidence with PET/MR was far higher than that with PET/CT. Similar results were found by Kuhn at el (12,13). This is likely due to the high soft-tissue contrast in MR compared with CT, which facilitates lesion characterization. Readers were less confident with bone lesions on PET/MR images, which is at variance with the known high confidence in high-spatial-resolution MR for neoplastic bone lesions. Reporting on incidental findings is an integral part of any imaging examination. Incidental findings may affect the course of disease, as they might be mistaken for malignancy or may require additional therapy. In total, PET/ CT demonstrated 17 findings more than PET/MR, potentially having an effect on patient management, but did not demonstrate six that were detected with PET/MR. This is partly in contrast
to the recent findings by Catalano et al, who reported that PET/MR has a higher clinical effect than PET/CT (14). However, they used a set of different MR protocols in their subjects, contrast material enhanced and nonenhanced, which were individually adapted to the patient’s disease. Unlike in our study, they acquired fully diagnostic MR images, generating an examination time of 66 minutes 6 12, whereas our PET/MR protocol was limited to the time range of a clinical PET/CT protocol. Our findings suggest that STIR PET/ MR is inferior to PET/CT, although STIR has been advocated for lung tumors in the pre-PET/MR era (9,15,16). LAVA PET/MR (ie, Dixon-type PET/MR) was only inferior to PET/CT in primary tumors, was less prone to artifacts than STIR, and is known to provide good anatomic localization (2,5). As the key questions asked in hybrid PET examinations concern lymph node and distant metastasis status, and since LAVA provides valuable information about tissue composition and seems to be a good sequence for MR-based attenuation correction in fully integrated systems, whole-body LAVA seems to be a valuable component of a PET/MR examination. Use of the PROPELLER sequence may provide an edge over PET/CT, even when used only in the chest and upper abdomen at whole-body PET/MR. Additional points, such as the rationale for the selection of MR pulse sequences, the conspicuity of small lesions, the issue with pulmonary nodules, and the significance of additional findings, are discussed in Appendix E5 (online). A limitation of our study is the inhomogeneous patient population regarding primary tumors and metastases, which obviates drawing conclusions about specific tumor entities. However, the goal of our study was to evaluate a number of MR sequences to design a valuable oncologic PET/MR protocol and to ensure that PET/MR and PET/ CT were comparable. A second limitation is that for ethical reasons, histologic findings were not available for all lymph nodes and distant metastases. Third, the analysis based on RECIST
radiology.rsna.org n Radiology: Volume 273: Number 3—December 2014
NUCLEAR MEDICINE: Whole-Body Nonenhanced PET/MR versus PET/CT in Cancer Staging
target lesion selection criteria might have introduced bias. However, this approach ensured that the most important lesions for oncologic staging and restaging were assessed. Fourth, the study was performed without the use of contrast agents. However, contrast agents are not used consistently in PET/CT anyway, and we wanted to see whether a PET/MR protocol could be devised that matches the performance of a nonenhanced PET/CT protocol. Fifth, we also did not use MR-based attenuation correction for PET/MR, as CT-based attenuation correction was available. This drawback affects mostly therapy monitoring, where reliable attenuation correction is needed for comparative quantification. Furthermore, it can be assumed that the PET/MR attenuation correction issues will be overcome. In summary, our preliminary evaluation indicates that PET/MR with LAVA and PROPELLER performs comparably to PET/CT in whole-body oncologic imaging and neoplastic lung disease. The basic protocol proposed here, performed simultaneously with a PET/ MR system, would take no longer than PET/CT. Drawbacks of PET/MR are increased numbers of artifacts, which do not interfere with image interpretation, and fewer incidental nonneoplastic lung findings than with PET/CT, the detection of which has limited value and is frequently inconsequential. Further studies will be needed to evaluate whether selective PET/MR protocols with sequences tailored to certain tumors and body regions yield additional advantages of PET/MR over PET/CT without making PET/MR an excessively lengthy and therefore expensive imaging modality.
I.A.B. disclosed no relevant relationships. M.P. disclosed no relevant relationships. G.D. Activities related to the present article: author is an employee of GE Healthcare. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. G.K.v.S. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: author received research grants and payment from GE Healthcare for lecturing; author received payment from E. Lilly and Icon Medical Imaging for consulting. Other relationships: disclosed no relevant relationships. P.V.H. Activities related to the present article: institution received a grant from GE Healthcare. Activities not related to the present article: institution received grants from Siemens Medical Solutions, Philips Medical, and Roche Pharma; author received payment from GE Healthcare for speaking. Other relationships: disclosed no relevant relationships.
Disclosures of Conflicts of Interest: M.W.H. Activities related to the present article: institution received a grant from GE Healthcare. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. P.A. disclosed no relevant relationships. F.P.K. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: author received payment for travel/accommodations/ meeting expenses from GE Healthcare Switzerland. Other relationships: disclosed no relevant relationships. L.H. disclosed no relevant relationships. C.M.P. disclosed no relevant relationships.
6. Eisenhauer EA, Therasse P, Bogaerts J, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45(2):228–247.
References 1. Appenzeller P, Mader C, Huellner MW, et al. PET/CT versus body coil PET/MRI: how low can you go? Insights Imaging 2013;4(4): 481–490. 2. Drzezga A, Souvatzoglou M, Eiber M, et al. First clinical experience with integrated whole-body PET/MR: comparison to PET/ CT in patients with oncologic diagnoses. J Nucl Med 2012;53(6):845–855. 3. Yi CA, Shin KM, Lee KS, et al. Non-small cell lung cancer staging: efficacy comparison of integrated PET/CT versus 3.0-T wholebody MR imaging. Radiology 2008;248(2): 632–642. 4. Stolzmann P, Veit-Haibach P, Chuck N, et al. Detection rate, location, and size of pulmonary nodules in trimodality PET/CT-MR: comparison of low-dose CT and Dixon-based MR imaging. Invest Radiol 2013;48(5): 241–246. 5. Eiber M, Martinez-Möller A, Souvatzoglou M, et al. Value of a Dixon-based MR/PET attenuation correction sequence for the localization and evaluation of PET-positive lesions. Eur J Nucl Med Mol Imaging 2011;38(9): 1691–1701.
7. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33(1):159–174. 8. Wiesmüller M, Quick HH, Navalpakkam B, et al. Comparison of lesion detection and quantitation of tracer uptake between PET from a simultaneously acquiring wholebody PET/MR hybrid scanner and PET
Radiology: Volume 273: Number 3—December 2014 n radiology.rsna.org
Huellner et al
from PET/CT. Eur J Nucl Med Mol Imaging 2013;40(1):12–21. 9. Ohno Y, Koyama H, Nogami M, et al. STIR turbo SE MR imaging vs. coregistered FDGPET/CT: quantitative and qualitative assessment of N-stage in non-small-cell lung cancer patients. J Magn Reson Imaging 2007; 26(4):1071–1080. 10. Chandarana H, Heacock L, Rakheja R, et al. Pulmonary nodules in patients with primary malignancy: comparison of hybrid PET/MR and PET/CT imaging. Radiology 2013;268(3):874–881. 11. Both M, Schultze J, Reuter M, et al. Fast T1- and T2-weighted pulmonary MR-imaging in patients with bronchial carcinoma. Eur J Radiol 2005;53(3):478–488. 12. Kuhn FP, Hüllner M, Mader CE, et al. Contrast-enhanced PET/MR imaging versus contrast-enhanced PET/CT in head and neck cancer: how much MR information is needed? J Nucl Med 2014;55(4):551–558. 13. Kuhn FP, Crook DW, Mader CE, Appen zeller P, von Schulthess GK, Schmid DT. Discrimination and anatomical mapping of PET-positive lesions: comparison of CT attenuation-corrected PET images with coregistered MR and CT images in the abdomen. Eur J Nucl Med Mol Imaging 2013;40(1): 44–51. 14. Catalano OA, Rosen BR, Sahani DV, et al. Clinical impact of PET/MR imaging in patients with cancer undergoing same-day PET/CT: initial experience in 134 patients— a hypothesis-generating exploratory study. Radiology 2013;269(3):857–869. 15. Ohno Y, Koyama H, Onishi Y, et al. Nonsmall cell lung cancer: whole-body MR examination for M-stage assessment—utility for whole-body diffusion-weighted imaging compared with integrated FDG PET/CT. Radiology 2008;248(2):643–654. 16. Morikawa M, Demura Y, Ishizaki T, et al. The effectiveness of 18F-FDG PET/CT combined with STIR MRI for diagnosing nodal involvement in the thorax. J Nucl Med 2009; 50(1):81–87. 17. Ma J. Breath-hold water and fat imaging using a dual-echo two-point Dixon technique with an efficient and robust phase-correction algorithm. Magn Reson Med 2004; 52(2):415–419. 18. Pipe JG. Motion correction with PROPELLER MRI: application to head motion and free-breathing cardiac imaging. Magn Reson Med 1999;42(5):963–969.
869
