Sebastian T. Schindera, MD Brian J. Soher, PhD David M. Delong, PhD Brian M. Dale, PhD Elmar M. Merkle, MD
1 From the Department of Radiology, Duke University Medical Center, Duke North, Room 1417, Erwin Rd, Durham, NC 27710 (S.T.S., B.J.S., D.M.D., E.M.M.); and Siemens Medical Solutions, Cary, NC (B.M.D.). Received June 18, 2007; revision requested August 27; revision received October 20; accepted January 7, 2008; final version accepted January 8. Address correspondence to E.M.M. (e-mail: elmar.merkle@duke.edu).
@ RSNA, 2008
Effect of Echo Time Pair Selection on Quantitative Analysis for Adrenal Tumor Characterization with In-Phase and Opposed-Phase MR Imaging: Initial Experience1
| Purpose: | To determine the effect of two pairs of echo times (TEs) for in-phase (IP) and opposed-phase (OP) 3.0-T magnetic resonance (MR) imaging on (a) quantitative analysis pro- spectively in a phantom study and (b) diagnostic accuracy retrospectively in a clinical study of adrenal tumors, with use of various reference standards in the clinical study. |
| Materials and Methods: | A fat-saline phantom was used to perform IP and OP 3.0-T MR imaging for various fat fractions. The institutional review board approved this HIPAA-compliant study, with waiver of informed consent. Single-breath-hold IP and OP 3.0-T MR images in 21 patients (14 women, seven men; mean age, 63 years) with 23 adrenal tumors (16 adeno- mas, six metastases, one adrenocortical carcinoma) were reviewed. The MR protocol involved two acquisition schemes: In scheme A, the first OP echo (approximately 1.5-msec TE) and the second IP echo (approximately 4.9- msec TE) were acquired. In scheme B, the first IP echo (approximately 2.4-msec TE) and the third OP echo (ap- proximately 5.8-msec TE) were acquired. Quantitative analysis was performed, and analysis of variance was used to test for differences between adenomas and nonadeno- mas. |
| Results: | In the phantom study, scheme B did not enable discrimina- tion among voxels that had small amounts of fat. In the clinical study, no overlap in signal intensity (SI) index values between adenomas and nonadenomas was seen (P < . 05) with scheme A. However, with scheme B, no overlap in the adrenal gland SI-to-liver SI ratio between adenomas and nonadenomas was seen (P < . 05). With scheme B, no overlap in adrenal gland SI index-to-liver SI index ratio between adenomas and nonadenomas was seen (P < . 05). |
| Conclusion: | This initial experience indicates SI index is the most reli- able parameter for characterization of adrenal tumors with 3.0-T MR imaging when obtaining OP echo before IP echo. When acquiring IP echo before OP echo, however, nonadenomas can be mistaken as adenomas with use of the SI index value. @ RSNA, 2008 |
P recise characterization of adrenal tumors is necessary to plan an ap- propriate clinical approach. In- phase (IP) and opposed-phase (OP) magnetic resonance (MR) imaging has proved highly accurate in the differenti- ation of adenomas from nonadenomas (1-4). The diagnostic efficiency of IP and OP MR imaging is due to its high sensitivity in distinguishing tissues with different lipid contents by using the in- terference of fat and water signals caused by the change in chemical shift frequency shift (5). In contrast to non- adenomas, adrenal adenomas generally contain variable amounts of intracyto- plasmatic lipid that result in signal in- tensity (SI) loss at OP imaging (6,7). To quantify this SI loss, various formulas used to calculate the relative change in SI have been published. The two most reliable formulas have proved to be those used to calculate the adrenal gland SI-to-spleen SI ratio (1,3,6,8).
At 1.5 T, the frequency shift be- tween water and fat signals is approxi- mately 225 Hz, resulting in IP signals at echo times (TEs) of 0.0 msec, 4.4 msec, 8.8 msec, and so on, and OP signals at TEs of 2.2 msec, 6.6 msec, 11.0 msec, and so on. At 3.0 T, the frequency shift between water and fat is doubled to ap- proximately 450 Hz, thus halving the 1.5-T IP and OP TEs. The 1.1-msec in- terval between IP and OP TEs makes collecting these echoes right after the other technically challenging. Currently, acquiring the first OP echo with a TE of 1.1 msec and the first IP echo only 1.1 msec later and within the same breath hold is not feasible with standard imag- ing sequences. Thus, two alternate data
Advances in Knowledge
At 3.0 T, the selection of echo pairs for in-phase (IP) and op- posed-phase (OP) MR imaging greatly affects the quantitative analysis used to differentiate ad- renal adenomas from nonadeno- mas.
Signal intensity index is the most accurate measurement for quanti- tative analysis when the OP echo is acquired before the IP echo.
acquisition schemes have been imple- mented at 3.0 T: (a) collection of the first OP echo followed by collection of the second IP echo and (b) collection of the first IP echo followed by collection of the third OP echo (9-11).
The purpose of our study was to determine the effect of two pairs of TEs for IP and OP 3.0-T MR imaging on (a) quantitative analysis prospec- tively in a phantom study and (b) diag- nostic accuracy retrospectively in a clin- ical study of adrenal tumors, with use of various reference standards in the clini- cal study.
Materials and Methods
One author (B.M.D.) is an employee of Siemens Medical Solutions, the manu- facturer of two 3.0-T MR imagers used in our study. However, authors who were not consultants for Siemens Medi- cal Solutions had complete control of any data and information that might have presented a conflict of interest for this author.
Phantom Study
A 1.5-L cylindrical phantom was cre- ated with equal parts soybean oil (Spec- trum Naturals, Petaluma, Calif) and sa- line 0.9% (Hospira, Lake Forest, Ill) (Fig 1) (4). Images parallel to the fat- saline boundary were acquired with a standard dual-echo two-dimensional IP and OP fast low-angle shot sequence at both 3.0 T and 1.5 T (Magnetom Trio and Avanto, respectively; Siemens, Er- langen, Germany) by using a circular polarized head coil. The 3.0-T IP and OP MR sequence included two acquisi- tion schemes: (a) acquisition of the first OP echo (1.5 msec) and the second IP echo (4.9 msec) and (b) acquisition of the first IP echo (2.6 msec) and the third OP echo (6.2 msec). The following pa-
Implication for Patient Care
For IP and OP MR imaging at 3.0 T, selecting the correct quantita- tive analysis of the echo time pair can enable one to avoid misclassi- fication of nonadenomas as adre- nal adenomas.
rameters were used for both schemes: 111-msec repetition time, 180 × 180-mm field of view, 128 x 128 matrix, 890 Hz/pixel receiver bandwidth, and an- teroposterior phase-encoding direction. The 1.5-T IP and OP MR sequence was used to acquire the first OP and IP ech- oes (2.4 and 5.0 msec, respectively; 111- msec repetition time; 180 × 180-mm field of view; 128 × 128 matrix; 380 Hz/pixel and 420 Hz/pixel receiver bandwidths, respectively).
A research scientist (B.J.S.) with 17 years of MR imaging experience ob- tained mean SI values from centrally lo- cated 2 × 2 × 2-cm regions of interest. The SI index (SI]) was calculated with the following equation and plotted versus fat fraction: SI] = [(SIpp - SIOp)/SIIp] X 100%, where SIIp is SI for IP and Slop is SI for OP. SI index voxels were located to coincide with the point-resolved spa- tially localized spectroscopy (PRESS) voxels described later in this article.
Single-voxel PRESS MR spectroscopy data were co-located with transverse IP and OP images to calibrate maximum wa- ter-fat content and SI variations. Acquisi- tion parameters were as follows: 2000/30 (repetition time msec/TE msec), 1024 points, 1540-Hz sweep width, four signals acquired, and no water suppression. MR spectroscopy voxels were centered in the fat-saline phantom where the meniscus
Published online 10.1148/radiol.2481071069
Radiology 2008; 248:140-147
Abbreviations:
IP = in phase
OP = opposed phase
RCC = renal cell carcinoma
SI = signal intensity
TE = echo time
Author contributions:
Guarantors of integrity of entire study, S.T.S., E.M.M .; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, S.T.S .; clinical studies, B.M.D.,
E.M.M .; experimental studies, S.T.S., B.J.S., B.M.D.,
E.M.M .; statistical analysis, D.M.D .; and manuscript edit- ing, all authors
See Material and Methods for pertinent disclosures.
was flattest. We acquired 21 spectra, starting with the MR spectroscopy voxel entirely in the saline compartment. Be- cause of time constraints, the voxel was moved in 2-mm increments until it was entirely in the fat compartment. For each voxel location, integrated areas under the SI index magnitude plots were used to determine the ratio of water to fat (water,
Figure 1
Lipid - Soybean Oil
Saline (0.9%)
Figure 1: Schematic drawing of the fat-saline phantom consisting of 50% saline and 50% soy- bean oil. We acquired 41 transverse 20-mm-thick images parallel to the fat-saline boundary in 1-mm steps from -20 mm to 20 mm across the fat-sa- line boundary to capture images with SI variations due to the different ratios of fat and saline in the section. Data acquisition started in pure saline (-20 mm) and ended in pure fat (20 mm). The in-plane voxel dimensions were nominally 1.4 x 1.4 mm.
2.7-6.7 ppm; fat, 0.0-2.4 ppm). Abso- lute SIs from water-only voxels and from fat-only voxels were compared to esti- mate SI contributions from each compart- ment of the phantom. Water and fat ar- eas from all voxels were normalized to these maximum values to determine if signal contributions from each compart- ment changed linearly. The phantom was allowed to settle in the MR imager for 20 minutes prior to spectroscopy or mea- surement of TE.
Patients, Lesion Confirmation, and Reference Standards in the Clinical Study
This Health Insurance Portability and Accountability Act-compliant study was approved by our institutional review board, and the requirement for informed consent was waived. The institutional 3.0-T MR database was searched to iden- tify abdominal MR examinations per- formed between March 2004 and June 2007 with MR reports that included the following keywords: adrenal adenoma,
Figure 2
Patients with adrenal tumor underwent IP/OP MR imaging at 3 T n = 31 (33)
Excluded patients n = 10 (10):
· due to small size of lesion n= 4 (4)
· due to insufficient diagnostic proof n = 6 (6)
Final patient number n = 21 (23)
Patients underwent acquisition scheme A only n = 7 (7)
Patients underwent both acquisition schemes A and B n = 12 (14)
Patients underwent acquisition scheme B only n = 2 (2)
Figure 2: Flowchart provides information about the method of patient recruitment and the number of pa- tients who underwent the two IP and OP MR acquisition schemes. Data in parentheses indicate the number of lesions.
| Parameter | Magnetom Trio | Magnetom Tim Trio | Signa Excite | |||
|---|---|---|---|---|---|---|
| Scheme A | Scheme B | Scheme A | Scheme B | Scheme A | Scheme B | |
| Repetition time (msec) | 216 | 111 | 193 | 214 | 170 | 150 or 170 |
| TE (msec) | 1.5, 4.9 | 2.5, 6.2 | 1.6, 4.9 | 2.2, 5.7 | 1.5, 4.4 | 2.2 or 2.5, 5.5 or 5.8 |
| Flip angle (degrees) | 71 | 70 | 90 | 90 | 55 | 55 or 80 |
| Matrix | 144 × 256 | 180 × 320 | 205 × 256 | 192 × 256 | 192 × 320 | 192 × 320 or 160× 320 |
| Section thickness (mm) | 4-8 | 6 | 5 | 5.5 | 8 | 4-8 |
| No. of sections | 19-25 | 20 | 24 | 22 | 21 | 21-30 |
| No. of signals acquired | 1 | 1 | 1 | 1 | 1 | 1 |
| Field of view (cm) | ~26×35 | ~26×35 | 30 × 30 | 28 × 30 | 40 × 40 | 40 × 40 |
| No. of breath holds | 1 | 1 | 1 | 1 | 1 | 1 |
| Acquisition time (sec) | 18-21 | 20 | 22 | 23 | 19 | 19 |
Note .- In scheme A, the first OP echo (approximately 1.5-msec TE) and the second IP echo (approximately 4.9-msec TE) were acquired. In scheme B, the first IP echo (approximately 2.4-msec TE) and the third OP echo (approximately 5.8-msec TE) were acquired.
adrenal mass, adrenal tumor, and/or adrenal metastases. The review yielded 31 patients with 33 adrenal lesions (Fig 2). Adrenal lesions smaller than 1 cm in diameter were excluded because of potential difficulties with accurate region of interest placement. Lesions with insufficient diagnostic proof also were excluded. The final proof for di- agnosis of adrenal adenomas was based on at least one of the following three factors: pathology data; stability of the lesion size at follow-up com- puted tomography (CT), MR imaging, or both within at least 6 months; or lesion attenuation of less than 10 HU on unenhanced CT images (2,3). Final diagnosis of a nonadenoma was estab- lished on the basis of pathology data; an at least 30% change in the maximal diameter of the lesion at follow-up CT, MR imaging, or both within 6 months; or both.
Four patients were excluded because of the small size of the adrenal lesion (<1 cm), and six patients were excluded be-
Figure 3
SI Index (%)
50
Scheme B, 3.0 T
0
1.5 T Approach
-50
Scheme A, 3.0 T
-100
0
25
50
75
100
Fat Fraction (%)
cause of insufficient diagnostic proof. Thus, our final study group consisted of 21 patients (14 women, seven men; mean age, 63 years; age range, 26-88 years) with 23 adrenal tumors. Twelve of these patients underwent MR imaging for fur- ther diagnostic work-up of an adrenal mass detected at prior CT. The remaining nine patients underwent MR imaging of the kidneys (n = 4), liver (n = 3), and pancreas (n = 2).
In 15 patients, 16 tumors were diag- nosed as adrenal adenomas (mean max- imal diameter, 2.5 cm; range, 1.1-4.1 cm): Nine were diagnosed because they were stable in size at follow-up CT or MR imaging (mean follow-up period, 19 months; range, 6-44 months), five were diagnosed because unenhanced CT attenuation was less than 10 HU, and two were diagnosed at percutane- ous biopsy. Seven tumors in six patients were diagnosed as nonadenomas (mean diameter, 4.5 cm; range, 1.9-7.5 cm): There were four metastases from clear- cell renal cell carcinoma (RCC) in three patients, one metastasis from squamous cell carcinoma of the cervix, one metas- tasis from colon carcinoma, and one ad- renocortical carcinoma. Two metasta- ses from RCC in one patient were diag- nosed at follow-up CT within 3 months because of a marked decrease in the maximal diameters of the lesions after chemotherapy for RCC. The third me- tastasis from RCC was confirmed with pathologic analysis, and the fourth me- tastasis from RCC was confirmed on the basis of rapid enlargement at follow-up CT within 4 months in a patient with known RCC. The metastasis from squa- mous cell carcinoma of the cervix, the metastasis from colon carcinoma, and the adrenocortical carcinoma were con- firmed with pathologic analysis.
Clinical MR Imaging Technique
Clinical MR imaging was performed with three 3.0-T MR imagers-Magnetom Trio (Siemens) (13 patients), Magnetom Tim Trio (Siemens) (five patients), and Signa Excite (GE Healthcare, Milwaukee, Wis) (three patients)-and use of dedi- cated eight-channel torso-array receive- only coils. The adrenal MR imaging proto- col consisted of coronal half-Fourier
single-shot fast spin-echo T2-weighted images or single-shot fast spin-echo T2-weighted images and transverse T1-weighted dual-echo two-dimensional gradient-echo IP and OP images. The IP and OP images were acquired in the same breath hold with various MR pa- rameters (Table 1) and an anteroposte- rior phase-encoding direction. The IP and OP MR protocol included acquisi- tion with the following schemes: In scheme A, we acquired the first OP echo with a TE of 1.5-1.6 msec and the second IP echo with a TE of 4.4-4.9 msec. In scheme B, we acquired the first IP echo with a TE of 2.2-2.5 msec and the third OP echo with a TE of 5.5-6.2 msec.
In December 2005, a Signa Excite MR system and a Magnetom Tim Trio system were installed as second and third 3.0-T platforms in addition to the Magnetom Trio system already in use at our hospital. At that time, we realized that the vendors had implemented two echo pair schemes for IP and OP MR imaging on their platforms. Thus, we modified our 3.0-T adrenal pulse se- quence protocol to include both TE ac- quisition schemes in January 2006.
Acquisition with scheme A included 15 adenomas in 14 patients and six non- adenomas (three metastases from RCC, one metastasis from squamous cell car- cinoma of the cervix, one metastasis from colon cancer, and one adrenocor- tical carcinoma) in five patients (Fig 2). Acquisition with scheme B included nine adenomas in eight patients and seven nonadenomas in six patients. Eight adenomas in seven patients and six nonadenomas (three metastases from RCC, one metastasis from squa- mous cell carcinoma of the cervix, one metastasis from colon carcinoma, and one adrenocortical carcinoma) in five patients were examined with both scheme A and scheme B.
Clinical MR Image Analysis
A 4th-year radiology resident (S.T.S.) with 2 years of experience in body MR imaging performed quantitative analysis on an MR workstation (Leonardo; Sie- mens). The resident was blinded to the acquisition schemes. Region of interest
| Scheme and Lesion | SI Index (%) | Adrenal SI-to-Spleen SI Ratio (%) | Adrenal SI-to-Liver SI Ratio (%) | Adrenal SI-to-Muscle SI Ratio (%) |
|---|---|---|---|---|
| Scheme A | ||||
| Adenoma (n = 15) | 21.1 ± 12.2 (2.7-44.8) | -29.5 ± 12.7 (-50.1 to -9.3)* | -26.8 ± 13.4 (-48.7 to -2.1) | -9.3 ± 23.3 (-47.6 to 25.9) |
| Nonadenoma (n = 6) | -8.9 ± 5.7 (-13.6 to 0.7) | -17.9 ± 18.4 (-41.7 to -1.7)* | -7.7 ± 6.7 (-16.5 to 2.7) | 14.5 ± 17.1 (2.0-48.2) |
| Scheme B | ||||
| Adenoma (n = 9) | 37.4 ± 17.0 (15.5-63.8) | -29.7 ± 23.1 (-57.9-7.3) | -25.5 ± 19.0 (-55.1 to 3.6) | -18.4 ± 23.3 (-56.5 to 18.3) |
| Nonadenoma (n = 7) | 16.8 ± 6.9 (8.2-26.9) | 43.8 ± 47.3 (3.6-134.7) | 13.0 ± 6.1 (3.1-20.7) | 18.5 ± 29.2 (-2.5 to 82.9) |
Note .- Data are mean values ± standard deviations. Data in parentheses are ranges.
* P> .05.
values for SI measurements were ob- tained in the same section for the adre- nal mass and for three reference tissues (spleen, liver, and ipsilateral paraspinal muscle). Circular regions of interest were placed at homogeneous artifact- free areas in the center of the adrenal mass and were drawn as large as possi- ble without including the edge of the lesion. Circular regions of interest in the spleen, liver, and paraspinal muscle were drawn as large as possible (range, 400-2000 mm2) without including blood vessels or fat tissue. By applying the copy and paste function of the worksta- tion, identical region of interest posi- tions could be achieved on correspond- ing IP and OP images. The SI value of each adrenal mass and the three refer- ence tissues was measured three times on the IP and OP images. With use of mean SI values and the following equa- tions, we calculated (a) SI index values, [(SIIp - SIOp)/(SIIp)] × 100%; (b) the ad- renal SI-to-spleen SI ratio, [(SIOpA/ SIOps)/(SIIp4/SIpps)-1] × 100%, where SlopA is the adrenal SI for OP, SIOps is the spleen SI for OP, SIIpA is the adrenal SI for IP, and SIIps is the spleen SI for IP; (c) the adrenal SI-to-liver SI ratio, [(SIOpA/
SIOpt)/(SI[p4/SIjpL.)- 1] × 100%, where Slopl is the liver SI for OP and SIIpl is the liver SI for IP; and (d) the adrenal SI-to- muscle SI ratio, [(SIOpA/SIOpM)/(SIIpA/ SIppM)-1] × 100%, where SlopM is the muscle SI for OP and SIIpM is the muscle SI for IP (1,6,8).
Clinical Data Evaluation, Statistical Analysis, and Reference Standards
The SI index and SI ratio for adenomas and nonadenomas were graphically de-
picted as scatterplots for the two acquisi- tion schemes. Preliminary data for scheme A indicated mean SI index values of approximately 20% for adenomas and -10% for nonadenomas, with a standard deviation of approximately 10 both for adenomas and for nonadenomas. With use of these values, a power analysis en- abled us to determine that a minimal sam- ple size of five adenomas and two tumors would be sufficient to obtain a statistical power of .8150 for the unpaired Student t test on the mean. With each acquisition protocol and evaluation method, a two- sample t test (equal variances not as- sumed) was used to test for mean differ- ences between adenomas and nonadeno- mas. A repeated-measures analysis of variance was used to test for mean differ- ences between scheme A and scheme B on the subset of lesions (n = 14) that were examined by using both acquisition schemes. Additionally, a receiver operat- ing characteristic curve was generated for each protocol and evaluation method. The area under the curve for each evalua- tion method, a suggested threshold, and corresponding sensitivity and specificity ratings were determined. A P value of less than .05 was considered to indicate a sig- nificant difference. The Holm-Bonferroni method was used to adjust for multiple comparisons (12). All statistical analyses were performed with SPSS 13.0 software (SPSS, Chicago, Ill).
Results
Phantom Study
As the MR section was moved through the phantom from pure fat to pure wa-
ter, MR spectroscopy peak areas for the 21 spectra acquired enabled us to con- firm approximately equal maximum wa- ter and fat signal contributions and a linear variation of these signals in the center of the phantom. The MR spec- trum of the soybean oil had a large peak at 440 Hz from the water peak (65% total area). Smaller peaks at 490 Hz, 370 Hz, and 246 Hz away from water contributed the remaining area. Over- all, the maximum signal measured from the fat region was 85% of that of the maximum water region signal.
There was a marked difference be- tween the in vitro SI index plots for scheme A and scheme B acquired at 3.0 T (Fig 3). The shape of the plot for scheme A at 3.0 T was similar to that at 1.5 T.
Clinical Study
No overlap was seen in the scatterplot for the SI index values with scheme A or for the adrenal SI-to-liver SI ratio with scheme B (Table 2; Figs 4a, 4c). How- ever, the scatterplot for the adrenal-to- muscle ratio and the adrenal-to-spleen ratio showed an overlap for schemes A and B, the scatterplot for the adrenal SI-to-liver SI lesion ratio showed an overlap for scheme A, and the scatter- plot for the SI index showed an overlap for scheme B (Fig 4). The adenomas were significantly different from the nonadenomas for all evaluation meth- ods and acquisition schemes (P < . 05), with the exception of the adrenal-to- spleen ratio with scheme A (Table 2). Scheme A was significantly different from scheme B (P < . 05) for the SI index and the adrenal SI-to-liver SI ra-
tio. Five of the six nonadenomas that had been imaged with both scheme A and scheme B showed a negative SI in- dex value with scheme A and a positive SI index value with scheme B (Table 2; Figs 4a, 5).
With acquisition scheme A, the SI index had an area under the curve of 1.000 (Table 3), which, when we used a threshold of 1.7% to distinguish adeno- mas from nonadenomas, resulted in sensitivity and specificity of 100%. With acquisition scheme B, the SI index per- formance was markedly degraded, with an area under the curve of only 0.889. When the data were collected with scheme B, the adrenal SI-to-liver SI ra- tio was the best measure (area under the
curve, 1.000), with a threshold of -0.2%, resulting in sensitivity and specificity of 100%.
Discussion
There are fundamental differences in the MR physics of 3.0-T units and 1.5-T units; therefore, some of the insights gained for IP and OP MR imaging at 1.5 T cannot be applied to higher-field- strength MR imaging. At 3.0 T, the TE pairs for IP and OP MR imaging need to be adjusted because the frequency dif- ference is double that of standard 1.5-T MR systems (13). With a 3.0-T MR sys- tem, fat and water protons are OP rela- tive to each other at 1.1 msec, 3.3 msec,
Figure 4
Scheme A
Scheme B
Scheme A
Scheme B
80
150
x
:
SI Index (%)
Adrenal-to-Spleen Ratio (%)
60
100
40
50
₹
xxx xxx
x
20
0
₹
*
…
0
xxxx x
-50
*
*
-20
-100
A
NA
A
NA
A
NA
A
NA
a.
b.
40
Scheme A
Scheme B
100
Scheme A
Scheme B
Adrenal-to-Liver Ratio (%)
80
×
20
xxx ** x
Adrenal-to-Muscle Ratio (%)
60
×
40
0
x xxxx xx 2
20
xxxxx
-20
0
-20
-40
-40
*
-60
-60
-80
A
NA
A
NA
A
NA
A
NA
C.
d.
Figure 4: Scatterplot shows the values of the (a) SI index, (b) adrenal SI-to-spleen SI ratio, (c) adrenal SI-to-liver SI ratio, and (d) adrenal SI-to-muscle SI ratio for the adenomas (A) and nonadenomas (NA) ex- amined with TE acquisition schemes A and B. There was no overlap between SI index values of the adenomas and nonadenomas for scheme A; however, a substantial overlap was found for scheme B. When scheme B was used, no overlap occurred between the adrenal SI-to-liver SI ratio values of the adenomas and nonadenomas; however, a considerable overlap was seen for scheme A.
5.5 msec, and so on, and IP relative to each other at 2.2 msec, 4.4 msec, 6.6 msec, and so on. Acquisition of the first OP echo at 1.1 msec and the first IP echo at 2.2 msec within the same breath hold would require unacceptably high receiver bandwidths. The most obvious solution is to acquire the first OP echo and the first IP echo during different breath holds. However, this approach can make quantitative analysis of partic- ularly smaller adrenal lesions less reli- able due to differences in section selec- tion. Thus, all three major MR vendors (GE Healthcare, Philips, and Siemens) currently recommend the collection of either the first OP signal and the second IP signal or the first IP signal and the third OP signal at 3.0-T MR imaging (9,10).
When we obtained the first OP echo before the second IP echo, we did not observe any overlap of SI index values between the 15 adenomas and the six nonadenomas. In contrast, acquisition of the first IP image before the third OP image resulted in a substantial overlap of SI index values between the adeno- mas and nonadenomas. In two studies in which a shorter TE was used for the IP echo compared with that used for the OP echo at magnetic field strengths lower than 3.0 T, researchers found similar overlaps of SI index values be- tween adenomas and malignant lesions (14,15). In these studies, several malig- nant tumors had a misleading SI loss on the OP images, as did seven nonadeno- mas in our study.
At IP and OP MR imaging, two fac- tors influence SI loss on OP images: chemical shift effect and T2* decay. The latter is the most plausible explanation for the misleading SI loss associated with nonadenomas with little to no lipid content on OP images when the IP echo is acquired before the OP echo. This scenario should yield a positive SI index value, whereas reverse TE selection is most likely to yield a negative SI index value. This pattern was demonstrated with our fat-saline phantom, in which the 3.0-T plot for the SI index of scheme B never reached negative values. In our clinical study, we found five nonadeno- mas that had been examined with both
scheme A and scheme B, which demon- strated a negative SI index value and a positive SI index value, respectively. Thus, radiologists selecting a TE that is shorter for the IP echo than for the OP echo are confronted with a diagnostic dilemma since they are unable to decide whether the SI loss in an adrenal lesion on an OP image is caused by chemical shift effect in a fat-containing adenoma or by T2* decay in a malignant lesion with little to no fat. In 1995, Tsushima and Dean (16) noted the influence of the selection of TEs on the SI index for IP and OP MR imaging when characteriz- ing adrenal tumors. Despite this alert, until 2006, various investigators still ac- quired the IP echo before the OP echo for both 1.5- and 3.0-T chemical shift MR imaging (7,10,11,15).
Some authorities have suggested ap- plying an internal reference tissue to re- duce the effect of T2* decay (1,7,17,18). Reference tissues used in the past have included the spleen, liver, and paraspi- nal muscle. In several investigations, researchers have recommended the spleen as the most suitable reference tissue since fat infiltration of the liver or paraspinal muscle may substantially influence any quantitative evaluation (1,7,17). However, the SI ratio of the spleen can also be subject to alteration, namely in the case of abnormal iron deposition (eg, hemosiderosis). In our study, we found the adrenal SI-to-liver SI ratio obtained by using scheme B to be an excellent discriminator between adrenal adenomas and nonadenomas, whereas the adrenal-to-spleen ratio demonstrated some overlap. Thus, ra- diologists who acquire the IP echo be- fore the OP echo and who use the liver or spleen as an internal reference tissue need to be cautious and aware of he- patic steatosis and splenic iron deposi- tion.
A limitation of our investigation was the small number of adrenal tumors that were included. However, despite the small number of lesions, our study had sufficient statistical power, as indicated by both the power analysis and the sig- nificant differences for most combina- tions of our evaluation methods and ac- quisition schemes. To further address
Figure 5
a.
b.
C.
d.
Table 3
Receiver Operating Characteristic Curve Analysis, Suggested Threshold, Sensitivity, and Specificity of the Four Evaluation Methods or Acquisition Schemes A and B
| Scheme and Evaluation Method | Area under the Receiver Operating Characteristic Curve | Suggested Threshold (%) | Sensitivity (%) | Specificity (%) |
|---|---|---|---|---|
| Scheme A | ||||
| SI index | 1.000* | 1.7 | 100 | 100 |
| Adrenal-to-spleen ratio | 0.689 | -13.0 | 67 | 87 |
| Adrenal-to-liver ratio | 0.867* | -18.1 | 100 | 80 |
| Adrenal-to-muscle ratio | 0.778 | 0.9 | 100 | 73 |
| Scheme B | ||||
| SI index | 0.889* | 20.2 | 71 | 89 |
| Adrenal-to-spleen ratio | 0.968* | -0.8 | 100 | 89 |
| Adrenal-to-liver ratio | 1.000* | -0.2 | 100 | 100 |
| Adrenal-to-muscle ratio | 0.857* | -5.1 | 100 | 78 |
* P> .05.
this limitation, we added a phantom study that clearly demonstrated the ef- fect of T2* decay on the quantitative analysis of IP and OP MR images at 3.0
T. On the basis of our findings and the fact that, to our knowledge, no scientific report had been published on IP and OP 3.0-T MR imaging for characterization
of adrenal lesions, we believe it is ap- propriate to inform the radiologic com- munity of the effect of T2* decay on IP and OP MR imaging and to provide pre- liminary guidelines for IP and OP 3.0-T MR imaging in the characterization of adrenal lesions. Another limitation of our study was that four metastases were from clear-cell RCC. It is well known that this tumor type may con- tain small amounts of intracytoplas- matic fat (19,20). However, in our study, the metastases from clear-cell RCC did not indicate an SI loss on OP images obtained with scheme A. Finally, only two echo pairs for IP and OP MR imaging were evaluated in our study. In future studies, a systematic comparison between multiple echo pairs, including a sequence with shortest IP and OP echo, should be performed to determine the optimal echo pairing.
In conclusion, selection of the TE scheme for IP and OP 3.0-T MR imaging may greatly affect the quantitative anal- ysis of adrenal tumors. To avoid mis- classification of adrenal tumors on IP and OP MR images, the SI index should be applied when the TE of the IP echo is longer than the TE of the OP echo. If the IP echo is collected before the OP echo, however, we recommend the use of an internal reference tissue, such as the liver or spleen, to correct for T2* decay effects.
Acknowledgments: We thank Richard Young- blood, MA, for editing the manuscript and Lisa K. Wall, BSRT, for searching the institutional 3.0-T MR database.
References
1. Fujiyoshi F, Nakajo M, Fukukura Y, Tsuchimochi S. Characterization of adrenal tumors by chemical shift fast low-angle shot
MR imaging: comparison of four methods of quantitative evaluation. AJR Am J Roentge- nol 2003;180:1649-1657.
2. 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.
3. Israel GM, Korobkin M, Wang C, Hecht EN, Krinsky GA. Comparison of unenhanced CT and chemical shift MRI in evaluating lipid- rich adrenal adenomas. AJR Am J Roentge- nol 2004;183:215-219.
4. Namimoto T, Yamashita Y, Mitsuzaki K, et al. Adrenal masses: quantification of fat content with double-echo chemical shift in- phase and opposed-phase FLASH MR im- ages for differentiation of adrenal adenomas. Radiology 2001;218:642-646.
5. Mitchell DG, Kim I, Chang TS, et al. Fatty liver: chemical shift phase-difference and suppression magnetic resonance imaging techniques in animals, phantoms, and hu- mans. Invest Radiol 1991;26:1041-1052.
6. Tsushima Y, Ishizaka H, Matsumoto M. Ad- renal masses: differentiation with chemical shift, fast low-angle shot MR imaging. Radi- ology 1993;186:705-709.
7. Mayo-Smith WW, Lee MJ, McNicholas MM, Hahn PF, Boland GW, Saini S. Characteriza- tion of adrenal masses (< 5 cm) by use of chemical shift MR imaging: observer perfor- mance versus quantitative measures. AJR Am J Roentgenol 1995;165:91-95.
8. Hussain HK, Korobkin M. MR imaging of the adrenal glands. Magn Reson Imaging Clin N Am 2004;12:515-544, vii.
9. Merkle EM, Nelson RC. Dual gradient-echo in-phase and opposed-phase hepatic MR imaging: a useful tool for evaluating more than fatty infiltration or fatty sparing. Radio- Graphics 2006;26:1409-1418.
10. von Falkenhausen MM, Lutterbey G, Morakkabati-Spitz N, et al. High-field- strength MR imaging of the liver at 3.0 T: intraindividual comparative study with MR imaging at 1.5 T. Radiology 2006;241:156- 166.
11. Mckenzie CA, Sacchetti D, Fredericks S, et al. Comparison of in and out of phase imaging and IDEAL fat-water separation for liver imaging at 3.0T [abstr]. In: Proceedings of the 14th Meeting of the International Soci- ety for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 2006; 631.
12. Holm S. A simple sequentially rejective multi- ple test procedure. Scand J Stat 1979;6:65-70.
13. Merkle EM, Dale BM. Abdominal MRI at 3.0 T: the basics revisited. AJR Am J Roentgenol 2006;186:1524-1532.
14. Reinig JW, Stutley JE, Leonhardt CM, Spicer KM, Margolis M, Caldwell CB. Differ- entiation of adrenal masses with MR imaging: comparison of techniques. Radiol- ogy 1994;192:41-46.
15. Slapa RZ, Jakubowski W, Dabrowska E, et al. Magnetic resonance imaging differenti- ation of adrenal masses at 1.5 T: T2- weighted images, chemical shift imaging, and Gd-DTPA dynamic studies. MAGMA 1996;4:163-179.
16. Tsushima Y, Dean PB. Characterization of adrenal masses with chemical shift MR imaging: how to select echo times. Radiology 1995;195:285-286.
17. Bilbey JH, McLoughlin RF, Kurkjian PS, et al. MR imaging of adrenal masses: value of chemical-shift imaging for distinguishing ad- enomas from other tumors. AJR Am J Roentgenol 1995;164:637-642.
18. Schwartz LH, Panicek DM, Koutcher JA, et al. Adrenal masses in patients with malignancy: prospective comparison of echo-planar, fast spin-echo, and chemical shift MR imaging. Radiology 1995;197:421- 425.
19. Shinozaki K, Yoshimitsu K, Honda H, et al. Metastatic adrenal tumor from clear-cell re- nal cell carcinoma: a pitfall of chemical shift MR imaging. Abdom Imaging 2001;26:439- 442.
20. Outwater EK, Bhatia M, Siegelman ES, Burke MA, Mitchell DG. Lipid in renal clear cell carcinoma: detection on opposed-phase gradi- ent-echo MR images. Radiology 1997; 205: 103-107.