GENITOURINARY IMAGING
ORIGINAL RESEARCH
Juliano F. Faria, MD Suzan M. Goldman, MD, PHD Jacob Szejnfeld, MD, PhD Homero Melo, MSc Cláudio Kater, MD, PhD Philip Kenney, MD Martha P. Huayllas, MD, MSc Guilherme Demarchi, MD Viviane V. Francisco, MD Cássio Andreoni, MD, PHD Miguel Srougi, MD, PhD Valdemar Ortiz, MD, PhD Nitamar Abdalla, MD, PHD
1 From the Department of Diagnostic Imaging, Federal University of São Paulo, Napoleão de Barros, 800, Vila Clementino, São Paulo, SP, Brazil 04024-002. From the 2005 RSNA Annual Meeting. Received October 28, 2006; revision requested January 10, 2007; revision received March 6; accepted April 11; final version accepted April 18. Address correspondence to J.F.F. (e-mail: drjulianounifesp@hotmail.com).
@ RSNA, 2007
Adrenal Masses: Characterization with in Vivo Proton MR Spectroscopy-Initial Experience1
| Purpose: | To prospectively determine the accuracy of in vivo proton (1H) magnetic resonance (MR) spectroscopy in distin- guishing adrenal adenomas, pheochromocytomas, adreno- cortical carcinomas, and metastases, with histologic or computed tomographic findings and follow-up data as the reference standards. |
| Materials and Methods: | This study was approved by the institutional ethics com- mittee, and informed consent was obtained. Sixty consec- utive patients (24 male and 36 female patients; mean age, 53 years) harboring adrenal tumors larger than 2 cm in diameter (mean diameter, 4.6 cm ± 3.4 [standard devia- tion]) entered the study and were examined with a 1.5-T MR imaging system and point-resolved multivoxel 1H MR spectroscopy. Thirty-eight patients had adenomas; 10, pheochromocytomas; five, carcinomas; and seven, metas- tases. Amplitude values for choline, creatine, lipid, and a metabolite peak at precession frequency of 4.0-4.3 ppm were measured. Metabolite ratios (choline-creatine, cho- line-lipid, lipid-creatine, and 4.0-4.3 ppm/creatine) and cutoff values (obtained by using receiver operating charac- teristic analyses) were obtained and compared for each type of adrenal mass, which was identified previously on the basis of clinical, hormonal, and pathologic evidence. Results were evaluated with x2 and Student t tests. Signif- icance was inferred at P < . 05. |
| Results: | Cutoff values of 1.20 for the choline-creatine ratio (92% sensitivity, 96% specificity; P < . 01), 0.38 for the choline- lipid ratio (92% sensitivity, 90% specificity; P < . 01), and 2.10 for the lipid-creatine ratio (45% sensitivity, 100% specificity) enabled adenomas and pheochromocytomas to be distinguished from carcinomas and metastases. A 4.0- 4.3 ppm/creatine ratio greater than 1.50 enabled distinc- tion of pheochromocytomas and carcinomas from adeno- mas and metastases (87% sensitivity, 98% specificity; P < .01). The best distinction was obtained by comparing cho- line-creatine and 4.0-4.3 ppm/creatine ratios. |
| Conclusion: | 1H MR spectroscopy can be used to characterize adrenal masses on the basis of spectral findings for benign adeno- mas, carcinomas, pheochromocytomas, and metastases. @ RSNA, 2007 |
C omputed tomography (CT) and magnetic resonance (MR) imaging are well-established methods used to differentiate adenomas from nonade- nomatous adrenal lesions (1-3). Adre- nal imaging techniques include unen- hanced and contrast material-enhanced CT, MR imaging, and fluorine 18 fluoro- deoxyglucose positron emission tomog- raphy (PET) (4-9). However, difficul- ties associated with use of adrenal imag- ing remain not only for diagnosis of atypical adenomas but also for detection of other adrenal alterations, such as me- tastases, pheochromocytomas, and ad- renocortical carcinomas (10-13). The appearance of benign pheochromocyto- mas generally overlaps that of carcino- mas and metastases on CT and MR im- ages. There is no signal intensity de- crease on MR images, and benign pheochromocytomas have inconsistent, often extremely vascular behavior (14).
Proton (1H) MR spectroscopy is a noninvasive imaging technique that is used to measure the biochemical na- ture of living tissues. It can be per- formed with most 1.5-T MR imaging instruments, and the information it yields may represent additional data on tumor metabolism that can be use- ful for diagnosis. Several studies have been performed to evaluate 1H MR spectroscopy of brain tumors (15-23),
Advances in Knowledge
· For adrenal mass characteriza- tion, choline-creatine ratios greater than 1.20 yielded 92% sensitivity and 96% specificity, and choline-lipid ratios greater than 0.38 yielded 92% sensitivity and 90% specificity.
In the differentiation of carcino- mas and pheochromocytomas from adenomas and metastases, a 4.0-4.3 ppm/creatine ratio greater than 1.50 yielded 87% sensitivity and 98% specificity.
Use of both choline-creatine and 4.0-4.3 ppm/creatine ratios en- abled distinction between adeno- mas, pheochromocytomas, carci- nomas, and metastases in 54 of 60 lesions.
and this modality has been used to char- acterize hepatocellular, colorectal, breast, cervical, neck, ovarian, prostate, sali- vary gland, bone, and soft-tissue tumors (24-33).
There is no tumor-specific metabo- lite that can be detected with 1H MR spectroscopy (34). It is possible, how- ever, to detect specific patterns in changes in metabolite concentrations. Choline is a compound metabolite that participates in cellular membrane syn- thesis and breakdown and can be used as a tumor marker. The creatine level yields information on cell energy status, and this compound may be used as a standard metabolite. The importance of lipid peaks is uncertain because they are frequently associated with benign cellu- lar processes (eg, the intracellular lipid seen in benign adenomas).
To our knowledge, only one study has been performed to evaluate in vivo 1H MR spectroscopy for characteriza- tion of adrenal tumors (35). Adrenal cortical lesions differ in their lipid con- tent, and this difference can be ap- praised with 1H MR spectroscopy. Evi- dence from in vivo measurements is used to confirm that the percentage of lipid content is significantly lower in carcinomas than in adenomas. How- ever, the question remains: Is 1H MR spectroscopy able to generate new in- formation in addition to the high-quality anatomic data provided by conventional MR imaging, and could this information be used to replace or complement bi- opsy findings in the evaluation and clini- cal care of patients? Given the differ- ences in biochemical activity among ad- renal adenomas, pheochromocytomas, carcinomas, and metastases, 1H MR spectroscopy may be used to make this distinction. Thus, the purpose of our study was to prospectively determine the accuracy of in vivo 1H MR spectros- copy in distinguishing adrenal adeno-
Implication for Patient Care
· In vivo 1H MR spectroscopy en- ables noninvasive characterization of adrenal masses (adenomas, pheochromocytomas, carcinomas, and metastases).
mas, pheochromocytomas, adrenocor- tical carcinomas, and metastases, with histologic or CT findings and follow-up data as the reference standards.
Materials and Methods
Patients, Masses, and Reference Standards
This study was approved by the institu- tional ethics committee, and informed consent was obtained from all patients. Sixty-four consecutive patients (25 male and 39 female patients; mean age, 54 years) with adrenal masses who met our study criteria and had undergone previous imaging with either abdominal CT protocol (densitometry or washout) were prospectively evaluated with a dedicated 1H MR spectroscopy proto- col. The endocrinology and urology ser- vices referred patients to the depart- ment of diagnostic imaging between Au- gust 2004 and January 2006. We did not include patients with malignant tumors who already were involved in therapeu- tic chemotherapy protocols, those who had undergone previous adrenal biopsy, or those with an adrenal mass smaller than 2.0 cm in diameter. We excluded four adenomas with diameters of ap- proximately 2 cm since it was impossi- ble to obtain voxels eligible for analysis.
Published online 10.1148/radiol.2453061854
Radiology 2007; 245:788-797
Abbreviations:
PPV = positive predictive value RARE = rapid acquisition with relaxation enhancement
Author contributions:
Guarantors of integrity of entire study, J.F.F., S.M.G., J.S., H.M., C.K., P.K., M.P.H., V.V.F., C.A., N.A .; study con- cepts/study design or data acquisition or data analysis/ interpretation, all authors; manuscript drafting or manu- script revision for important intellectual content, all au- thors; manuscript final version approval, all authors; literature research, J.F.F., S.M.G., J.S., P.K., C.A .; clinical studies, J.F.F., S.M.G., J.S., M.P.H .; experimental studies, J.F.F., S.M.G., J.S., H.M., C.K., M.P.H., V.V.F., C.A., N.A .; statistical analysis, J.F.F., S.M.G., J.S., H.M., C.K., N.A .; and manuscript editing, J.F.F., S.M.G., J.S., H.M., C.K., P.K., M.P.H., G.D., V.V.F., M.S., V.O., N.A.
Authors stated no financial relationship to disclose.
Two patients had bilateral adrenal le- sions. In each case, only the larger le- sion was included in the study. Sixty patients (24 male and 36 female pa- tients; mean age, 53 years) harboring adrenal tumors larger than 2 cm in di- ameter (mean, 4.6 cm ± 3.4 [standard deviation]) entered the study (Fig 1).
We used a previously reported threshold (unenhanced CT attenua- tion ≤ 10 HU, enhancement washout ≥ 60%) as a reference standard to classify adrenal masses as adenomas. Criteria for adenoma diagnosis also included sta- ble size for at least 12 months and surgi- cal resection of functioning adenomas or atypical masses. Thirty-eight patients with adenomas were included in the study: 26 adenomas had an attenuation of no more than 10 HU on unenhanced CT images (-10 to 8 HU; mean, 3 HU), 24 had stable size (mean duration of follow- up, 22 months; range, 12-48 months), and two had been surgically resected (both had refractory hyperaldosteronism at presentation). Of the 38 adenomas, 12 had an attenuation of more than 10 HU at unenhanced CT (18-30 HU; mean, 23 HU). Two adenomas were surgically resected. At imaging, one patient had an adenoma with a mass diameter of 4 cm and washout of less than 60%, and the other patient had refractory hyper- aldosteronism. The other 10 adenomas had washout of at least 60% (range, 60%-86%; mean, 67%) and stable size (mean duration of follow-up, 25 months; range, 14-46 months).
Evaluation of pheochromocytomas and carcinomas included histopathologic analysis of surgical specimens (refer- ence standard). Five carcinomas and 10 pheochromocytomas were included. Eight pheochromocytomas were diagnosed biochemically, and two were not diag- nosed. Metastasis evaluation included histopathologic proof by means of bi- opsy or surgery. Seven metastases were included: surgically resected lung carci- nosarcoma (n = 1) and lung adenocarci- noma (n = 1) and biopsy-sampled lung carcinoma (n = 2), lymphoma (n = 1), germ cell carcinoma (n = 1), and mes- enchymal tumor (n = 1).
The mean diameters, with standard deviations, and diameter ranges (in pa-
rentheses) for the adenomas, pheochro- mocytomas, carcinomas, and metasta- ses were 2.8 cm ± 0.7 (2.1-4.4 cm), 5.5 cm ± 1.3 (3.4-8.4 cm), 10.1 cm ± 5.6 (4.7-18.4 cm), and 8.5 cm ± 4.7 (3.1-16.0 cm), respectively. Forty-eight masses were located in the right gland, and 12 were located in the left gland.
MR Imaging and 1H MR Spectroscopic Imaging
Examinations were performed with a 1.5-T (43 mT/m) MR unit (Sonata MC; Siemens, Erhlund, Germany) equipped with a phased-array coil. MR imaging performed at the level of the adrenal mass consisted of a T2-weighted se- quence and chemical shift imaging. Lo- calization images, which consisted of transverse, coronal, and sagittal T2- weighted sections, were obtained by us- ing a half-Fourier rapid acquisition with relaxation enhancement (RARE) se- quence with 4.4-msec echo space, rep- etition time msec/echo time msec of 900/90 (effective), 3-mm section thick- ness, 37 × 37-cm field of view, and 167 × 256 matrix. Localization sequences could be any breath-hold sequence with adequate time-space resolution.
Chemical shift MR imaging was per- formed by using a fast low-angle shot sequence with a repetition time of 100 msec (coronal plane) or 170 msec (transverse plane), double echo times of 2.4 and 4.8 msec, a 90° flip angle, 4-mm maximum section thickness, 167 x 256 matrix, 35 × 35-cm field of view, and one signal acquired. Half-Fourier RARE
sequences were performed in trans- verse, coronal, and sagittal planes for three-dimensional mass localization and 1H MR spectroscopy planning. To deter- mine the correct insertion for the vol- ume of interest, we used three localizing half-Fourier RARE sagittal sequences: at maximum inspiration, at maximum expiration, and at free breathing. In this way, we determined the range of posi- tioning, from the highest to the lowest point, where the gland could be located during the acquisition of 1H MR spectro- scopic images during free breathing. There would be a high probability that the adrenal gland and mass would be located in this interval. This allowed one of the authors (J.F.F., H.M .; each with 4 years of MR experience) to carefully position the multivoxel volume of inter- est grids in the center of the lesion, with use of all three sagittal sequences, to include as much of the lesion area as possible or, preferentially, to include all of the lesion and part of the adjacent fat tissue. The volume of interest grid was composed of 256 voxels, with a nominal voxel size of 0.75 × 0.75 × 1.0 cm (0.56 cm3), within a 16 × 12 × 0.75-cm field of view. The field homogeneity was opti- mized automatically over the selected grid of interest by observing the water signal intensity (Fig 2).
Three-dimensional 1H MR spectros- copy was performed by using a point-re- solved spatially localized spectroscopy se- quence with a two-dimensional chemical shift imaging sequence, multivoxel local- ization, 1500/135, and an acquisition time
Figure 1
Patients with adrenal masses n=64
Figure 1: Flowchart of study enrollment.
Excluded patients - no eligible voxels n=4
Index test n=60
No reference standard n=0
Adenoma n=38
Pheochromocytoma n=10
Carcinoma n=5
Metastases n=7
of approximately 6 minutes. Eight satura- tion region bands were used. Water sup- pression was achieved prior to the point- resolved spatially localized spectroscopic examination. One signal was acquired for each water suppression spectrum. The total amount of time needed for spectros- copy, including planning and data acquisi- tion, was, on average, 20 minutes.
1H MR Spectroscopic Data Analysis
After data acquisition, an author (J.F.F.) processed the 1H MR spectroscopy data
by using a protocol specifically designed for that purpose at a workstation (Leo- nardo; Siemens) with spectral analysis software, a 1000-Hz Gaussian line-broad- ening filter, and zero filling of Hanning (approximately 4-msec center and 200- msec width). Fourier transformation and automatic phase correction were also in- corporated in this order.
1H MR spectroscopy images were overlaid on the corresponding T2-weighted images and evaluated to determine which voxels were eligible for analysis. Individ-
Figure 2
1
2 Distance: 1.81 cm
1
2
2
3
A
B
10cm
C
ual voxels were considered eligible if they consisted of 100% lesion tissue. Voxels located in the adjacent fat were not in- cluded in the analysis. 1H MR spectros- copy images were interpreted by means of visual inspection and metabolite peak amplitude measurements. Analysis time varied according to the number of evalu- ated voxels and lasted up to 50 minutes. All spectral fits were performed in an analysis window from 0.50 to 5.0 ppm. Metabolites with a standard deviation of more than 20% were not included in the statistical analysis. Metabolite amplitude peaks from each eligible voxel were mea- sured and used to calculate the mean mass ratio. Lipid-positive peaks were de- termined in the chemical shift range of 0.90-2.02 ppm. Creatine- and choline- positive peaks were determined with chemical shifts of 3.08 and 3.22 ppm, re- spectively. All amplitudes were deter- mined for each metabolite of interest for every lesion; subsequently, all means, medians, and ranges were calculated for every metabolite ratio (choline to cre- atine, choline to lipid, and lipid to creat- ine). We noted a peak in the frequency range of 4.0-4.3 ppm, mainly in pheo- chromocytomas and carcinomas. A retro- spective evaluation was performed, and a 4.0-4.3 ppm/creatine ratio was calcu- lated.
Statistical Analysis
Differences in ratios between lesion types were correlated by using the x2 test with Yates correction or the Fisher exact test when a cell value was less than 5. Two- sample paired Student t tests were used to compare mean metabolite ratios for different mass groups. A P value of less than .05 indicated significance for all tests. Estimated power analysis for comparison of proportions was per- formed for all tables.
Receiver operating characteristics curve analyses were used for different ra- tio thresholds to ascertain the best cutoff values for sensitivity and specificity. Sen- sitivity, specificity, positive predictive value (PPV), and accuracy were deter- mined for each mass metabolite ratio cut- off value.
Analyses were conducted by using a software program (Epi Info, version
| Metabolite Ratio Values Obtained with 1H MR Spectroscopy | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Choline-Creatine Ratio | 4.0-4.3 ppm/Creatine Ratio | Lipid-Creatine Ratio | Choline-Lipid Ratio | |||||||||
| Adrenal Mass | Mean* | Median | Range | Mean* | Median | Range | Mean* | Median | Range | Mean* | Median | Range |
| Adenoma (n = 38) | 0.89 ± 0.17 | 1.0 | 0.47-1.0 | 0.77 ± 0.18 | 0.75 | 0.33-1.40 | 8.67 ± 4.83 | 7.0 | 3.0-23.0 | 0.15 ± 0.09 | 0.13 | 0.01-0.33 |
| Pheochromocytoma (n = 10) | 1.06 ± 0.26 | 1.0 | 0.80-1.75 | 3.10 ± 1.81 | 2.46 | 1.0-5.60 | 4.13 ± 3.60 | 2.85 | 1.0-12.0 | 0.47 ± 0.34 | 0.40 | 0.07-1.0 |
| Carcinoma (n = 5) | 1.62 ± 0.13 | 1.60 | 1.50-1.80 | 3.42 ± 2.10 | 4.0 | 0.80-5.55 | 3.62 ± 0.90 | 3.40 | 2.80-5.0 | 0.46 ± 0.09 | 0.41 | 0.38-0.60 |
| Metastases (n = 7) | 2.40 ± 0.90 | 2.50 | 1.27-4.0 | 1.35 ±0.36 | 1.40 | 0.90-2.0 | 1.50 ± 0.55 | 0.80 | 0.30-3.70 | 3.65 ± 2.10 | 2.20 | 0.70-13.0 |
| * Data are means ± standard deviations. | ||||||||||||
Figure 3: Point-resolved multivoxel 1H MR spectroscopy (1500/135) performed in (a) 48-year-old man with right adrenal lipid-rich adenoma and (b) 25-year-old man with left adrenal lipid-poor adenoma. Spectroscopy revealed an eligible internal voxel for each lesion. Reference images (1000/88) are located to the right of each spectrum and were acquired in the coronal (top), sagittal (middle), and transverse (bottom) planes in a and in the sagittal (top), coronal (middle), and transverse (bottom) planes in b. These images reveal the location of the analyzed voxel inside each lesion (voxel size, 0.56 cm3). The metabolite ratios are as follows: choline-creatine ratio, 0.76 in a and 0.69 in b; lipid-creatine ratio, 6.72 in a and 5.0 in b; choline-lipid ratio, 0.11 in a and 0.14 in b; and 4.0-4.3 ppm/creatine ratio, 0.71 in a and 0.68 in b. A = amplitude, AU = arbitrary units, Cho = choline, Cr = creatine, and LIP = lipid.
Figure 3
LIP
LIP
A: 0.794
A: 0.219
0.2
0.5
Intensity (AU)
Cho
A: 0.0912
Intensity (AU)
4.0-4.3 ppm
Cr
0.1
A: 0.0846
A: 0.119
Cr
4.0-4.3 ppm
Cho
A: 0.044V
0.0
A: 0.0301
A: 0.0305
0.0
-0.5
5
4
3
2
1
5
4
3
2
1
ppm
ppm
a.
b.
3.3.2, 2005; Centers for Disease Control and Prevention, Atlanta, Ga). Receiver operating characteristics curve analysis was performed with statistical software (SPSS for Windows, version 10; SPSS, Chicago, Ill). Power analysis was per- formed with other software (Stata, ver- sion 8.2; Stata, College Station, Tex).
Results
Eligible Voxels
On average, there were 9 eligible voxels (range, 3-52 voxels) in adenomas, 46
(range, 16-144 voxels) in pheochromo- cytomas, 76 (range, 48-144 voxels) in carcinomas, and 39 (range, 14-144 voxels) in metastases.
Visual Inspection and Metabolite Ratios
We observed different spectral pat- terns for different types of adrenal masses (Table 1). Adenomas had rela- tively homogeneous spectra, with low variability among each of the eligible voxels, and, in general, they had only positive lipid peaks in the spectra. We did not find significant differences (P >
.05) in metabolite peaks between lipid- rich and lipid-poor adenomas (Fig 3). Conversely, pheochromocytomas, car- cinomas, and metastases commonly re- vealed marked spectral variability at vi- sual inspection (Figs 4-7).
Metastases had the most heteroge- neous spectra, with many different me- tabolite peaks, a predominance of posi- tive choline peaks, and no positive lipid peaks. Carcinomas had high variability of spectra within the mass. Positive lipid peaks were seen frequently, while posi- tive choline peaks were present in only
a few (at least four) eligible voxels. Pos- itive 4.0-4.3-ppm peaks were also present. There was little variability in the spectra of pheochromocytomas, with positive lipid and 4.0-4.3-ppm peaks. Few positive choline peaks were de- tected (less than two spectra). There were particular voxels in some of the lesions that had lipid and 4.0-4.3-ppm peaks with too wide a base, and these
were confused with the adjacent cre- atine and choline peaks.
The results enabled us to differentiate the masses and group them (Tables 2-5). Carcinomas and metastases had choline- creatine ratios greater than 1.20 in 11 (92%) of 12 masses, while adenomas and pheochromocytomas had choline-cre- atine ratios equal to or less than 1.20 in 46 of 48 masses (92% sensitivity, 96%
specificity, 86% PPV, 95% accuracy, P < . 01). The areas under the receiver operating characteristics curve for the choline-creatine ratio and 4.0-4.3 ppm/creatine ratio were 0.91 (95% confidence interval: 0.81, 1.01) and 0.97 (95% confidence interval: 0.92, 1.01), respectively. High-grade tumors had higher choline-creatine ratios than did low-grade tumors. A choline-lipid
Figures 4-6
4.0-4.3 ppm
A: 1.03
1.0
Intensity (AU)
Cho
A: 0.541
0.5
Cr
0.382
LIP
A: 0.206
0.0
5
4
3
2
1
ppm
4.
Cho A: 0,0678
0.05
4.0-4.3 ppm
A: 0.0141
Intensity (AU)
0.0164
LIP
0.00481
0.00
-0.05
5
4
3
2
1
ppm
6.
LIP
A: 0.275
4.0-4.3 ppm
A: 0.219
0.2
Intensity (AU)
Cho
0.1
A: 0.0758
Cr
A: 0.0398
0.0
-0.1
5
4
3
2
1
ppm
5.
ratio greater then 0.38 enabled dis- tinction of carcinomas and metastases from pheochromocytomas or adeno- mas (92% sensitivity, 90% specificity, 69% PPV, 90% accuracy, P < . 01). A lipid-creatine ratio of less than 2.10 en- abled clear distinction of carcinomas, metastases, and pheochromocytomas from adenomas (45% sensitivity, 100% specificity, 100% PPV, 80% accuracy). There were no significant differences (P > .05) in the lipid-creatine ratios between lipid-rich and lipid-poor adeno- mas.
A 4.0-4.3 ppm/creatine ratio greater than 1.50 enabled distinction of carcino- mas and pheochromocytomas from me- tastases and adenomas (87% sensitiv- ity, 98% specificity, 98% PPV, 95% accuracy, P < . 01). A 4.0-4.3 ppm/ creatine ratio greater than 1.50 proved to be useful for differentiating pheo- chromocytomas from adenomas (80% sensitivity, 100% specificity). In our sample, only three adenomas had rare 4.0-4.3-ppm peaks in eligible voxels. However, discrimination was possible even in these cases, since the average 4.0-4.3 ppm/creatine ratio in the eligi- ble voxels was still lower than 1.50. There were no significant differences (P > .05) in the 4.0-4.3 ppm/creatine ratios between pheochromocytomas and carcinomas. Estimated power for two-sample comparison of proportions for all results in the tables was 1.0. Al- though these relationships are more in- formative than the mean metabolite val- ues, only groups of tissues could be dif- ferentiated. Consequently, we decided to compare the information from the two most informative relations. This comparison, performed with use of a scatterplot, allowed us to differentiate the four tissue types in 54 of 60 patients (Fig 8).
Discussion
With use of 1H MR spectroscopy visual inspection, important differences that enabled us to distinguish benign from malignant adrenal masses were ob- served only in the presence of a positive choline peak. However, the accuracy of
1.0
1.0
0.8-
0.8
Sensitivity
0.6-
Sensitivity
0.6
0.4
0.4
0.2
0.2
0.0
0.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
1 - Specificity
1 - Specificity
a.
b.
Figure 7: Receiver operating characteristic curves. (a) 1H MR spectroscopy choline-creatine ratio for use in differentiation of carcinomas and metastases versus adenomas and pheochromocytomas. The area under the receiver operating characteristics curve was 0.91 (95% confidence interval, 0.81, 1.01). When the choline- creatine ratio was greater than 1.20, the best combination of sensitivity (92%) and specificity (96%) was pro- vided. (b) 1H MR spectroscopy 4.0-4.3 ppm/creatine ratio for use in differentiation of carcinomas and pheo- chromocytomas versus adenomas and metastases. The area under the receiver operating characteristics curve was 0.97 (95% confidence interval: 0.92, 1.01). When the 4.0-4.3 ppm/creatine ratio was greater than 1.50, the best combination of sensitivity (87%) and specificity (98%) was provided.
| Choline-Creatine Ratio | No. of Carcinomas and Metastases | No. of Adenomas and Pheochromocytomas |
|---|---|---|
| >1.20 | 11 | 2 |
| ≤1.20 | 1 | 46 |
Note .- Sensitivity, 92%; specificity, 96%; PPV, 85%; accuracy, 95%; P < . 01 (Fisher exact test).
| Choline-Lipid Ratio | No. of Carcinomas and Metastases | No. of Adenomas and Pheochromocytomas |
|---|---|---|
| >0.38 | 11 | 5 |
| ≤0.38 | 1 | 43 |
Note .- Sensitivity, 92%; specificity, 90%; PPV, 69%; accuracy, 90%; P < . 01 (Fisher exact test).
1H MR spectroscopy visual inspection was worse than that of 1H MR spectros- copy metabolite ratio analyses, which enabled distinction of benign adenomas, pheochromocytomas, carcinomas, and metastases.
In our sample, the choline-creatine ratio was the most useful metabolite ra- tio for differentiation of metastases and carcinomas from pheochromocytomas and adenomas, and the 4.0-4.3 ppm/ creatine ratio was best for differentia-
tion of pheochromocytomas and carci- nomas from metastases and adenomas. It was useful to include the other ratios, even though there was some overlap in the differentiation between pheochro- mocytomas, adenomas, and carcinomas.
Lipid-rich adenomas contain more intracytoplasmic lipids than do lipid- poor adenomas; however, lipid-creatine ratios of lipid-poor and lipid-rich adeno- mas were not significantly different. Lipid-creatine ratios had low sensitivity but high specificity and PPV for ade- noma diagnosis. The 4.0-4.3 ppm/cre- atine ratios of carcinomas and pheo- chromocytomas were not significantly different, but the choline-creatine ratio helped with the differentiation of them. Because of the frequent presence of a 4.0-4.3 ppm positive peak in pheochro- mocytomas and carcinomas, we de- cided to use this metabolite to create a new ratio: 4.0-4.3 ppm/creatine. To our surprise, this ratio turned out to be one of the most useful for differentia- tion. To our knowledge, this peak still has not been completely described and defined in the literature, but it may rep- resent blood breakdown products (36). Rarely, adenomas may contain hemor- rhage foci (37). Small pheochromocyto- mas tend to be solid, whereas hemor-
rhagic and cystic areas become more common with increasing size (38).
Evidence that high-grade tumors had higher choline-creatine ratios than did low-grade tumors has also been found in studies on malignant central nervous system and neck tumors (39,40). The usefulness of information generated from cell proliferation studies per- formed with healthy and diseased adre- nal tissues is still controversial. A group of authors (41) found no significant dif- ference in proliferation fractions be- tween benign lesions and carcinomas. We noticed significant differences in the choline-creatine ratios between benign and malignant adrenal lesions; how- ever, in carcinomas, the average limits were close to those of adenomas and pheochromocytomas. Because of the marked heterogeneity, the procedure used to detect a positive choline peak in large carcinomas is not simple. This is not true for adrenal metastases, as pos- itive choline peaks frequently were found in eligible voxels, particularly in high- grade lesions.
There were limitations to our study. Lesions smaller than 2 cm in diameter are generally unsuitable for analysis be- cause of the absence of eligible voxels. This is due to artifacts caused by respi-
ratory movements, which do not allow insertion of the volume of interest in the three acquisitions in the sagittal plane-at inspiration and expiration and during free breathing. Use of 1H MR spectroscopy sequences during apnea could reduce this problem. Use of higher-field-strength MR imagers allows acquisition of reliable spectra from smaller lesions with faster imaging times and makes differentiation between the various metabolite peaks studied more accurate; further investi- gation of our technique with such in- struments would be useful (42). Also, development of specific software for use with this analysis method would help to simplify measurements. Another prob- lem was the possibility that a curve made at the edge of the lesion could have been contaminated by adjacent fat. However, we tested several periph- eral noneligible voxels, compared them with spectra that were considered eligi- ble, and found important differences in the curves. Our findings about the use- fulness of this technique need to be val- idated by the findings of further studies. Such studies should include additional types of adrenal lesions, including nodu- lar hyperplasia, myelolipomas, cysts, hematomas, pediatric neoplasms, and inflammatory processes. The informa-
| Lipid-Creatine Ratio | and Pheochromocytomas | No. of Adenomas |
|---|---|---|
| ≤2.10 | 10 | 0 |
| >2.10 | 12 | 38 |
Note .- Sensitivity, 45%; specificity, 100%; PPV, 100%; accuracy, 80%.
| 4.0-4.3 ppm/Creatine Ratio | No. of Pheochromocytomas and Carcinomas | No. of Adenomas and Metastases |
|---|---|---|
| >1.50 | 13 | 1 |
| ≤1.50 | 2 | 44 |
Note .- Sensitivity, 87%; specificity, 98%; PPV, 98%; accuracy, 95%; P < . 01 (Fisher exact test).
≥3
Choline/creatine ratio
2
1.2
1
0
0
1
1.5
2
≥4.0
4.0-4.3ppm/creatine ratio
Figure 8: Two-dimensional scatterplot based on the results obtained with the receiver operating characteristic curves for choline-creatine and 4.0-4.3 ppm/creatine ratios. Simultaneous use of these analyses for each type of adrenal mass en- abled 54 of 60 masses to be sorted into four dis- tinct groups. O = adenoma, = pheochromo- cytoma, = carcinoma, and= metastases.
tion provided by 1H MR spectroscopy could complement the findings of other modalities, such as CT, MR imaging, and fluorodeoxyglucose PET.
In conclusion, multivoxel 1H MR spectroscopy can be used to character- ize and distinguish the various adrenal masses, yielding different spectral find- ings for adenomas, pheochromocyto- mas, carcinomas, and metastases. Cho- line-creatine ratios greater than 1.20 yielded 92% sensitivity and 96% speci- ficity, and choline-lipid ratios greater than 0.38 yielded 92% sensitivity and 90% specificity; both were sufficient for differentiation of adenomas and pheo- chromocytomas from carcinomas and metastases. In the differentiation of car- cinomas and pheochromocytomas from adenomas and metastases, a 4.0-4.3 ppm/creatine ratio greater than 1.50 yielded 87% sensitivity and 98% speci- ficity. Simultaneous use of these two analyses for each type of adrenal mass made it possible to sort 54 of 60 masses into four distinct groups.
References
1. Caoili EM, Korobkin M, Francis IR, Cohan RH, Dunnick NR. Delayed enhanced CT of lipid poor adrenal adenomas. AJR Am J Roentgenol 2000;175:1411-1415.
2. Caoili EM, Korobkin M, Francis IR, et al. Adrenal masses: characterization with com- bined unenhanced and delayed enhanced CT. Radiology 2002;222:629-633.
3. Rescinito G, Zandrino F, Cittadini G Jr, Santacroce E, Giasotto V, Neumaier CE. Characterization of adrenal adenomas and metastases: correlation between non-en- hanced computed tomography and chemical shift magnetic resonance imaging. Acta Ra- diol 2006;47:71-76.
4. Goldman SM, Coelho RD, Freire Filho Ede O, et al. Imaging procedures in adrenal pa- thology. Arq Bras Endocrinol Metabol 2004; 48:592-611.
5. Maurea S, Mainolfi C, Bazzicalupo L, et al. Imaging of adrenal tumors using FDG PET: comparison of benign and malignant lesions. AJR Am J Roentgenol 1999;173:25-29.
6. Mackie GC, Shulkin BL, Ribeiro RC, et al. Use of 18F-deoxyfluoroglucose positron emission tomography in evaluating locally re- current and metastatic adrenocortical carci-
noma. J Clin Endocrinol Metab 2006;91: 2665-2671.
7. Haider MA, Ghai S, Jhaveri K, Lockwood G. Chemical shift MR imaging of hyperattenuat- ing (>10 HU) adrenal masses: does it still have a role? Radiology 2004;231:711-716.
8. Israel GM, Korobkin M, Wang C, Hecht EN, Krinsky GA. Comparison of non-enhanced CT and chemical shift MRI in evaluating lipid-rich adrenal adenomas. AJR Am J Roentgenol 2004;183:215-219.
9. Savci G, Yazici Z, Sahin N, Akgoz S, Tuncel E. Value of chemical shift subtraction MRI in characterization of adrenal masses. AJR Am J Roentgenol 2006;186:130-135.
10. Paulsen SD, Nghiem HV, Korobkin M, Caoilli EM, Higgins EJ. Changing role of im- aging guided percutaneous biopsy of adrenal masses: evaluation of 50 adrenal biopsies. AJR Am J Roentgenol 2004;182:1033-1037.
11. Park BK, Kim B, Ko K, Jeong SY, Kwon GY. Adrenal masses falsely diagnosed as adeno- mas on non-enhanced and delayed contrast- enhanced computed tomography: pathologi- cal correlation. Eur Radiol 2006;16:642- 647.
12. Blake MA, Krishnamoorthy SK, Boland GW, et al. Low-density pheochromocytoma on CT: a mimicker of adrenal adenoma. AJR Am J Roentgenol 2003;181:1663-1668.
13. Blake MA, Kalra MK, Sweeney AT, et al. Distinguishing benign from malignant adre- nal masses: multi-detector row CT protocol with 10-minute delay. Radiology 2006;238: 578-585.
14. Szolar DH, Korobkin M, Reittner P, et al. Adrenocortical carcinomas and adrenal pheochromocytomas: mass and enhance- ment loss evaluation at delayed contrast-en- hanced CT. Radiology 2005;234:479-485.
15. Rutter A, Hugenholtz H, Saunders JK, Smith IC. Classification of brain tumors by ex vivo 1H NMR spectroscopy. J Neurochem 1995; 64:1655-1661.
16. Rand SD, Prost R, Haughton V, et al. Accu- racy of single voxel proton MR spectroscopy in distinguishing neoplastic from nonneo- plastic brain lesions. AJNR Am J Neurora- diol 1997;18:1695-1704.
17. Bitsch A, Brhun H, Vougioukas V, et al. In- flammatory CNS demyelination: histopatho- logic correlation with in vivo quantitative proton spectroscopy. AJNR Am J Neurora- diol 1999;20:1619-1627.
18. Majós C, Alonso J, Aguilera C, et al. Adult primitive neuroectodermal tumor: proton MR spectroscopic findings with possible ap- plication for differential diagnosis. Radiology 2002;225:556-566.
19. Kugel H, Heindel W, Ernestus RI, Bunke J, Mesnil R, Friedmann G. Human brain tumors: spectral patterns detected with lo- calized 1H-MR spectroscopy. Radiology 1992;183:701-709.
20. Poptani H, Gupta RK, Roy R, Pandey R, Jain VK, Chhabra DK. Characterization of intra- cranial mass lesions with in vivo proton MR spectroscopy. AJNR Am J Neuroradiol 1995; 16:1593-1603.
21. Adamson AJ, Rand SD, Prost RW, Kim TA, Schultz C, Haughton VM. Focal brain lesions: effect of single voxel proton MR spectroscopic findings on treatment deci- sions. Radiology 1998;209:73-78.
22. Meyerand ME, Pipas JM, Mamourian A, Tosteson TD, Dunn JF. Classification of bi- opsy confirmed brain tumors using single- voxel MR spectroscopy. AJNR Am J Neuro- radiol 1999;20:117-123.
23. Stadlbauer A, Gruber S, Nimsky C, et al. Preoperative grading of gliomas by using me- tabolite quantification with high-spatial-reso- lution proton MR spectroscopic imaging. Ra- diology 2006;238:958-969.
24. Swindle P, McCredie S, Russell P, et al. Pathologic characterization of human pros- tate tissue with proton MR spectroscopy. Ra- diology 2003;228:144-151.
25. Mackinnon WB, Huschtscha L, Dent K, Hancock R, Paraskeva C, Mountford CE. Correlation of cellular differentiation in hu- man colorectal carcinoma and adenoma cell lines with metabolite profiles determined by 1H magnetic resonance spectroscopy. Int J Cancer 1994;59:248-261.
26. Wang CK, Li CW, Hsieh TJ, Chien SH, Liu GC, Tsai KB. Characterization of bone and soft-tissue tumors with in vivo 1H-MR spectroscopy: initial results. Radiology 2004; 232:599-605.
27. Schellinger D, Lin CS, Fertikh D, et al. Nor- mal lumbar vertebrae: anatomic, age, and sex variance in subjects at proton MR spec- troscopy-initial experience. Radiology 2000;215:910-916.
28. Singer S. New diagnostic modalities in soft tissue sarcoma. Semin Surg Oncol 1999;17: 11-22.
29. Mukherji SK, Schiro S, Castillo M, et al. Proton MR spectroscopy of squamous cell carcinoma of the upper aerodigestive tract: in vitro characteristics. AJNR Am J Neurora- diol 1996;17:1485-1490.
30. Yeung DK, Cheung HS, Tse GM. Human breast lesions: characterization with con- trast-enhanced in vivo proton MR spectros- copy-initial results. Radiology 2001;220: 40-46.
31. Maheshwari SR, Mukherji SK, Neelon B, et al. The choline creatine ratio in five benign neoplasms: comparison with squamous cell carcinoma by use of in vitro MR spectros- copy. AJNR Am J Neuroradiol 2000;21: 1930-1935.
32. Lim AK, Patel N, Gedroyc WM, Blomley MJ, Hamilton G, Taylor-Robinson SD. Hepato- cellular adenoma: diagnostic difficulties and novel imaging techniques. Br J Radiol 2002; 75:695-699.
33. King AD, Yeung DK, Ahuja AT, et al. Sali- vary gland tumors at in vivo proton MR spec- troscopy. Radiology 2005;237:563-569.
34. Hom JJ, Coakley FV, Simko JP, et al. Pros- tate cancer: endorectal MR imaging and MR spectroscopic imaging-distinct true posi- tive results from chance detected lesions. Radiology 2006;238:192-199.
35. Leroy-Willig A, Bittoun J, Luton JP, et al. In vivo MR spectroscopic imaging of adrenal glands: distinction between adenomas and carcinomas larger than 15 mm based on lipid content. AJR Am J Roentgenol 1989;153: 771-773.
36. Allen JR, Prost RW, Griffith OW, Erickson SJ, Erickson AB. In vivo proton (H1) mag- netic resonance spectroscopy for cervical carcinoma. Am J Clin Oncol 2001;24:522- 529.
37. Elsayes KM, Mukundan G, Narra VR, et al. Adrenal masses: MR imaging features with pathologic correlation. RadioGraphics 2004; 24(suppl 1):S73-S86.
38. Blake MA, Kalra MK, Maher MM, et al. Pheochromocytoma: an imaging chameleon. RadioGraphics 2004;24(suppl 1):S87-S99.
39. Magalhaes A, Godfrey W, Shen Y, Hu J, Smith W. Proton magnetic resonance spec- troscopy of brain tumors correlated with pa- thology. Acad Radiol 2005;12:51-57.
40. King AD, Yeung DK, Ahuja AT, Leung SF, Tse GM, van Hasselt AC. In vivo proton MR spectroscopy of primary and nodal nasopha- ryngeal carcinoma. AJNR Am J Neuroradiol 2004;25:484-490.
41. Wajchenberg BL, Pereira MA, Medonca BB, et al. Adrenocortical carcinoma. Cancer 2000;88:711-736.
42. Gonen O, Gruber S, Li BS, Mlynárik V, Moser E. Multivoxel 3D proton spectroscopy in the brain at 1.5 versus 3.0 T: signal to noise ratio and resolution comparison. AJNR Am J Neuroradiol 2001;22:1727-1731.