3D Turbo Spin-Echo Sequence with Motion-Sensitized Driven-Equilibrium Preparation for Detection of Brain Metastases on 3T MR Imaging

E. Nagao; T. Yoshiura; A. Hiwatashi; M. Obara; K. Yamashita; H. Kamano; Y. Takayama; K. Kobayashi; H. Honda

doi:10.3174/ajnr.A2343

Abstract

BACKGROUND AND PURPOSE: MSDE preparation is a technique for black-blood imaging. Our purpose was to evaluate the usefulness of a 3D TSE sequence with MSDE preparation in detecting brain metastases by comparing it with conventional sequences.

MATERIALS AND METHODS: Postcontrast images of 227 patients who were suspected of having brain metastasis were prospectively obtained by using 3 T1-weighted 3D sequences: a gradient-echo sequence (MPRAGE), TSE-noMSDE, and TSE-MSDE. The number of visualized blood vessels and the lesion-to-normal CNR were compared among the 3 sequences. An observer test involving 9 radiologists was performed, and their diagnostic performance by using TSE-MSDE, MPRAGE, and combined TSE-MSDE and MPRAGE was compared by means of an FOM as an index of diagnostic performance derived by the JAFROC analysis, sensitivity, FP/case, and reading time.

RESULTS: TSE-MSDE resulted in significantly better vessel suppression than the other 2 methods. TSE with and without MSDE resulted in significantly higher CNRs than MPRAGE. In the observer test, significantly higher sensitivity and FOM as well as significantly shorter reading time were achieved by TSE-MSDE compared with MPRAGE, but FP/case was significantly higher with TSE-MSDE. Combined TSE-MSDE/MPRAGE resulted in significantly higher sensitivity and FOM and similar FP/case and reading time compared with MPRAGE alone.

CONCLUSIONS: With blood vessel suppression and increased CNR, TSE-MSDE improves radiologists' performances in detecting brain metastases compared with MPRAGE, but it may increase FP results. Combined with MPRAGE, TSE-MSDE achieves high diagnostic performance while maintaining a low FP rate.

Abbreviations

ANOVA: analysis of variance
CNR: contrast-to-noise ratio
FA: flip angle
FOM: figure of merit
FP: false-positive
FP/case: false-positive results per case
G: gradient
GRE: gradient-recalled echo
JAFROC: Jackknife Free-Response Receiver Operating Characteristic
LCD: liquid crystal display
LSMeans: least square means
MPRAGE: magnetization-prepared rapid acquisition of gradient echo
MSDE: motion-sensitized driven-equilibrium
NS: not significant
RF: radio-frequency
TSE: turbo spin-echo
TSE-MSDE: TSE with MSDE preparation
TSE-noMSDE: TSE without MSDE preparation
VENC: velocity-encoding

Brain metastases are the most common intracranial tumors in adults, affecting ≤20% of patients with cancer.^1,2 They represent one of the most frequent neurologic complications of systemic cancer as a major cause of morbidity and mortality. Not only the presence or absence but also the number of brain metastases critically affects therapeutic strategies.^3,4 For example, in the management of patients with multiple brain metastases, stereotactic radiosurgery along with whole-brain radiation therapy has been recommended when the number of lesions is ≤4.^4,5 Thus, accurate pretreatment diagnosis by MR imaging is crucial.

Postcontrast 3D T1-weighted GRE imaging is widely used to detect brain metastases.^6,7 In these images, a high signal intensity of enhancing blood vessels is a disturbing factor because it can be mistaken for enhancing metastases. The suppression of blood vessel signals may simplify radiologists' reading processes and, in turn, improve their performance. In this study, we introduced an MR image with MSDE preparation,^8,9 aiming for blood vessel suppression. MSDE preparation is a technique for black-blood imaging and was first optimized to efficiently suppress signals from blood in imaging of the arterial wall.^8–10 With MSDE preparation, the flow-sensitizing gradient pair introduces the phase dispersion among moving spins, as opposed to the stationary spins, resulting in selective suppression of signals from flowing blood. We implemented an MSDE preparation with a T1-weighted 3D TSE sequence because it has recently been shown that 3D TSE imaging can increase the CNR for brain metastases compared with a conventional GRE sequence.^11,12 Therefore, our purpose was to evaluate the usefulness of TSE-MSDE in detecting brain metastases by comparison with conventional GRE imaging.

Materials and Methods

This study was approved by the institutional review board of our hospital. Written informed consent was obtained from volunteers and patients.

This study consisted of 3 parts. First, we performed a volunteer study to determine an appropriate VENC setting for TSE-MSDE. Next, we evaluated the image quality of the MSDE-TSE sequence by comparing it with conventional sequences. Specifically, the number of visualized vessels and lesion-to-normal CNR was compared. Finally, we conducted an observer test. For the latter 2 parts, we constructed a data base of patients' images from which appropriate cases were selected for the respective purposes.

Volunteer Study

With MSDE preparation, the degree of blood signal-intensity suppression depends on the VENC setting because lower VENC values result in the suppression of blood with a slower flow. 3D TSE-MSDE images of the brains of 2 healthy men (33 and 45 years old) were obtained by using a 3T MR imaging scanner (Achieva Quasar Dual; Philips Electronics, Best, the Netherlands) and an 8-channel array head coil after administration of a contrast agent (gadoteridol, ProHance; Eisai, Tokyo, Japan; 0.2 mmol/Kg). Figure 1 shows our TSE-MSDE sequence.¹³ Imaging parameters were as follows: TR/TE, 345/20 ms; FA, 90°; echo-train length, 11; refocus angle, 130°; sensitivity-encoding factors, 2.4 (phase) and 1.8 (section); FOV, 240 mm; voxel size, 1.0 × 1.0 × 1.0 mm³. Images were sequentially obtained at 6 VENC values: 7.2, 9.2, 10.8, 20.8, 39.6, and 55.0 mm/s. In our MSDE sequences, motion-sensitized gradients were inserted in all 3 directions. Therefore, the VENC values are indicated as resultant values. The first scan was started at 5 minutes after the contrast injection. To evaluate the possible effect of imaging order on the results, we reversed the order of the 6 scans in the second subject. The images were obtained in a sagittal plane, reconstructed into 2-mm-thick contiguous transverse sections, and displayed on an LCD monitor of a PACS (Rapideye; Toshiba Medical, Tokyo, Japan). For each VENC setting, blood vessels were visually counted by a board-certified radiologist (E.N.) in a single section at the level of the centrum semiovale.

Fig 1.

Diagram of a TSE-MSDE used in this study. The MSDE preparation consists of a 90° excitation pulse, two 180° Malcolm Levitt refocusing pulses with each pulse sandwiched by bipolar motion-sensitizing gradients, and a −90° flip back pulse. The motion-sensitized gradients are inserted in all 3 (x, y, and z) directions. All refocusing pulses are implemented as composite pulses (90x-180y-90x). Bipolar gradients are used to reduce eddy current effects. An additional bipolar gradient is inserted in front of the sequence for further eddy current compensation.

Data Base of Patient Images

We constructed a data base of patient images by prospectively imaging consecutive patients who were referred for MR evaluation for possible brain metastasis. A total of 259 head MR imaging examinations from 227 patients were included in the data base. The MR imaging hardware and the contrast agent were the same as those used for the volunteer study. For each patient, we obtained 3 types of 3D postcontrast T1-weighted images sequentially: GRE (MPRAGE), TSE-noMSDE, and TSE-MSDE sequences. The parameters of the TSE-MSDE sequence are shown in the previous section. VENC was set at an appropriate value determined in the volunteer study, and the resultant shortest preparation duration was 14.7 ms. The parameters for MPRAGE were TR/TE, 8.2/3.8 ms; TI, 1028 ms; FA, 8°; sensitivity-encoding factors, 1 (phase) and 2.5 (section). The parameters for TSE-noMSDE were exactly the same as those for TSE-MSDE except for the absence of MSDE preparation. Imaging times for MPRAGE, TSE-noMSDE, and TSE-MSDE were 5 minutes 20 seconds, 7 minutes, and 7 minutes, respectively. All 3 sequences were obtained in a sagittal plane with the same FOV (240 mm) and voxel size (1.0 × 1.0 × 1.0 mm³) and were reconstructed in 2-mm-thick contiguous transverse images. The first postcontrast scan was started at 5 minutes after the contrast injection. The imaging order was MPRAGE, TSE-noMSDE, and TSE-MSDE (order 1 hereafter) in the first 62 examinations. The reverse order (order 2) was used for the remaining 197 examinations.

Determination of Metastatic Lesions

From the image data base, patients with a history of pathologically proved cancer and with at least 1 follow-up MR imaging examination were selected. There were 132 such patients. Two board-certified radiologists (E.N., T.Y.) reviewed all images of these patients in a consensus reading and selected intraparenchymal enhancing lesions that were visualized in both MPRAGE and TSE-MSDE. We excluded lesions visualized on only 1 of the 2 sequences, to stay conservative in determining the criterion standard. An effort was made to exclude artifacts and enhancing vessels, by carefully comparing the 2 images. Among the selected lesions, those satisfying any of the following conditions were determined to be metastases: 1) the size of the lesion increased in the follow-up period, 2) the lesion newly appeared during the follow-up period, and 3) the lesion received a treatment (either systemic chemotherapy or brain radiation therapy). Four hundred eighty-five lesions of 31 patients were recognized as metastases. During the same review, 174 patients were found to have no enhancing lesions.

Evaluation of Image Quality

Fifty patients with no enhancing lesion (35 men, 15 women; age, 26–87 years), including 25 imaged in order 1 and 25 in order 2, were randomly selected from the image data base. The images of these patients were further evaluated on the PACS by 1 board-certified radiologist (E.N.). For each of the 3 imaging methods, the degree of blood vessel suppression was evaluated by counting the visualized blood vessels in a single section at the level of the centrum semiovale. In addition, the lesion-to-normal CNR was compared. For the CNR evaluation, we selected lesions with a diameter ≥ 5 mm and homogeneous solid enhancement. As a result, 29 metastatic lesions, including 11 imaged in order 1 (1 man, 2 women; age, 46–74 years; 2 patients with lung cancer and 1 with thyroid cancer) and 18 in order 2 (8 male patients with lung cancer; age, 47–77 years) were selected. The CNR was calculated as follows: where SI_lesion and SI_background represent the mean signal intensities of the lesion and normal-appearing white matter in the same section, respectively, and SD_lesion and SD_background are the corresponding SDs.¹⁴ These parameters were obtained by defining circular regions of interest of 5 mm in each location.

Observer Study

Cases with 1–9 metastases were selected from the image data base, and as a result, 17 cases (12 men, 5 women; age, 37–87 years; 14 patients with lung cancer, 1 with oral floor carcinoma, 1 with breast cancer, and 1 with pancreatic cancer) with 50 lesions were chosen. Among the 17 cases, 1 case was imaged in order 1 and 16 in order 2. Figure 2 shows the distribution of lesion size. In addition, 17 cases without any enhancing lesions (12 men, 5 women; age, 48–84 years; 15 patients with lung cancer, 1 with gastric cancer, and 1 with renal cell carcinoma) were selected, among which 2 cases were scanned in order 1 and the other 15 in order 2. These 34 cases were used for the observer tests. We included only MPRAGE and TSE-MSDE for the observer test because an evaluation of image quality revealed that TSE-MSDE showed better blood vessel suppression than TSE-noMSDE and similar CNR.

Fig 2.

The distribution of the 50 metastatic lesions used in the observer study according to lesion diameter, which ranged from 2 to 36 mm, with an average of 6.8 mm.

Nine radiologists who were blinded to the patients' clinical information, including 4 first-year radiology residents who each had 6 months of experience in radiology practice and 5 board-certified radiologists who had 16, 13, 10, 9, and 7 years of experience, respectively, participated in the observer test. None of the 9 radiologists participated in the case selection. Each radiologist took part in 3 reading sessions, which were each at least 1 month apart. They read TSE-MSDE, MPRAGE, and both of these, in the first, second, and third reading sessions, respectively. All images were displayed on a 20.8-inch LCD monitor. The radiologists were allowed to freely increment the sections by using a mouse with a wheel. In each reading session, the 34 cases were presented in a randomized order. For a training session before the test, 3 training cases, each of which had 10–12 metastases, were provided to the radiologists to familiarize them with the operation of the PACS and the rating system. The radiologists were asked to place an arrow electronically in each location where they found a metastasis, to record the results of the readings. In the third reading session, in which both TSE-MSDE and MPRAGE were read, the observers were instructed to use TSE-MSDE to pick up high-signal-intensity areas as candidates for metastases and MPRAGE as a second opinion to reject FPs such as enhancing vessels and artifacts, taking advantage of the synchronized section increment capability of the PACS. The radiologists were asked to report their level of confidence in the presence of metastasis at each location by using a rating bar scale, on which the right and left ends corresponded to the highest (100) and lowest (0) confidence levels, respectively. The reading time for each case was recorded.

Statistical Analyses

The number of blood vessels and CNR was compared by using LSMeans Student t test when 2-way ANOVA was significant.

To evaluate the radiologists' performances in the observer test, we used JAFROC analysis.^15,16 This analysis has been proposed to statistically estimate the differences in diagnostic performance between different modalities when location issues are relevant. We applied JAFROC analysis with method 1 of Chakraborty and Berbaum¹⁵ to estimate an FOM value as an index of each radiologist's performance in each session. A free software package (JAFROC1, http://www.devchakraborty.com) was available. We also obtained the sensitivity and the number of FPs/case. Imaging findings related to the FP results judged by the consensus of the 2 board-certified radiologists (E.N., T.Y.) were reported. Additionally, sensitivities were compared among the 3 sessions according to lesion size: small (≤5 mm in shorter diameter) and large (>5 mm).

The sensitivity, FP/case, and reading time among the 3 reading sessions were compared by using a LSMeans Student t test when 2-way ANOVA was significant. The difference in sensitivity between small and large lesions was analyzed by using the t test. For all analyses, P < .05 was considered statistically significant.