Slip Interface Imaging Predicts Tumor-Brain Adhesion in Vestibular Schwannomas

Slip interface imaging, an MR elastographic imaging–based technique, provides a method to preoperatively determine the degree of tumor-brain adhesion in patients with vestibular schwannomas and offers a potential approach to improve preoperative planning, which includes determination of surgical risk and likelihood of gross total resection.

Abstract

Purpose

To test the clinical feasibility and usefulness of slip interface imaging (SII) to identify and quantify the degree of tumor-brain adhesion in patients with vestibular schwannomas.

Materials and Methods

With institutional review board approval and after obtaining written informed consent, SII examinations were performed in nine patients with vestibular schwannomas. During the SII acquisition, a low-amplitude mechanical vibration is applied to the head with a pillow-like device placed in the head coil and the resulting shear waves are imaged by using a phase-contrast pulse sequence with motion-encoding gradients synchronized with the applied vibration. Imaging was performed with a 3-T magnetic resonance (MR) system in less than 7 minutes. The acquired shear motion data were processed with two different algorithms (shear line analysis and calculation of octahedral shear strain [OSS]) to identify the degree of tumor-brain adhesion. Blinded to the SII results, neurosurgeons qualitatively assessed tumor adhesion at the time of tumor resection. Standard T2-weighted, fast imaging employing steady-state acquisition (FIESTA), and T2-weighted fluid-attenuated inversion recovery (FLAIR) imaging were reviewed to identify the presence of cerebral spinal fluid (CSF) clefts around the tumors. The performance of the use of the CSF cleft and SII to predict the degree of tumor adhesion was evaluated by using the κ coefficient and McNemar test.

Results

Among the nine patients, SII agreed with the intraoperative assessment of the degree of tumor adhesion in eight patients (88.9%; 95% confidence interval [CI]: 57%, 98%), with four of four, three of three, and one of two cases correctly predicted as no adhesion, partial adhesion, and complete adhesion, respectively. However, the T2-weighted, FIESTA, and T2-weighted FLAIR images that used the CSF cleft sign to predict adhesion agreed with surgical findings in only four cases (44.4% [four of nine]; 95% CI: 19%, 73%). The κ coefficients indicate good agreement (0.82 [95% CI: 0.5, 1]) for the SII prediction versus surgical findings, but only fair agreement (0.21 [95% CI: −0.21, 0.63]) between the CSF cleft prediction and surgical findings. However, the difference between the SII prediction and the CSF cleft prediction was not significant (P = .103; McNemar test), likely because of the small sample size in this study.

Conclusion

SII can be used to predict the degree of tumor-brain adhesion of vestibular schwannomas and may provide a method to improve preoperative planning and determination of surgical risk in these patients.

^© RSNA, 2015

Introduction

The goal of microsurgery for vestibular schwannomas is complete tumor removal with no new neurologic deficit. If the tumor is adherent to the brainstem, vasculature, or cranial nerves, these goals become more difficult. Less than gross total resection carries a higher risk of tumor remnant regrowth and the need for additional treatment with attendant morbidity (1–5). Also, a poor tumor-brain interface can prolong the surgical time, which increases the risk of postoperative complications, including infarction, deep venous thrombosis, or infection (6,7). Therefore, presurgical knowledge of tumor-brain adhesion would be helpful to predict potential complications, the length of surgery, and the likelihood of complete tumor resection.

A variety of imaging modalities attempted to help explore adhesion on the basis of tumor size, tumor signal intensity, the presence of a surrounding cerebrospinal fluid (CSF) cleft, peritumoral edema, and vascular supply (8–14). A magnetic resonance (MR) imaging–based method that involves imaging of two different phases of CSF pulsation was used to evaluate tumor-brain adhesion by assessing the dynamics of brain surface motion (15–17). However, the variations of pulsatile motion within the brain can compromise the sensitivity of this technique. Although these imaging methods led to some information about the presence of adhesions, they do not provide a direct measure of tumor-brain adhesion.

Shear stress causes two contiguous parts of a body to deform or slide relative to each other in a direction parallel to the applied stress. An adhesive interface that experiences a shear force will exhibit shear displacement continuity across the interface, while a nonadhesive interface may slip rather than deform, which causes a discontinuity in displacement. Recently, a shear line imaging method based on well-established MR elastography techniques was developed to assess this mechanical shear connectivity across tissue layers (18). It demonstrated the capability to view functional shear slip interfaces in phantoms and volunteer studies of the abdomen and forearms. In our study, we sought to apply shear line imaging to brain tumors to view tumor-brain adhesion. We also incorporated the calculation of octahedral shear strain (OSS; [19,20]) to quantify the degree of adhesion by assessing the change in shear displacement that occurs at the tumor-brain interface. With a three-dimensional displacement field, the OSS at a given voxel is the maximum change in shear displacement across all possible planes, and thus would be expected to be larger at slip interfaces than at adhesive interfaces. Because the extended technique focuses on making visible and quantifying the tumor-brain slip interface, we refer to this technique as slip interface imaging (SII).

We hypothesized that a nonadhesive tumor-brain interface would appear as a dark line on the shear line images because of intravoxel phase dispersion (IVPD) (18,21), and a nonadhesive interface would exhibit a larger OSS value than an adhesive interface. The purpose of this study was to test the clinical feasibility and usefulness of SII to identify and quantify the degree of tumor-brain adhesion in patients with vestibular schwannomas.

Materials and Methods

The Mayo Clinic and authors (K.J.G., A.M., R.L.E., and J.H.) have intellectual property rights and a financial interest through receipt of royalties and equity from licensing of MR elastography technology. R.L.E. is the chief executive officer of Resoundant (Rochester, Minn). This research was approved and conducted in compliance with oversight by the Mayo Clinic conflict of interest review board.

Patient Recruitment

This study was reviewed and approved by our institutional review board. Nine patients (seven women and two men; age range, 32–77 years) with presumed vestibular schwannomas on standard MR images and who were scheduled for resection were recruited for SII examinations. All patients provided written informed consent for the imaging study.

Surgical Evaluation of the Degree of Tumor-Brain Adhesion

At the time of surgery, two neurosurgeons (J.J.V.G., with 10 years of experience in neurosurgery, and M.J.L., with 18 years of experience in neurosurgery) who were blinded to the SII results and the neuroradiologist’s interpretation regarding tumor-brain adhesion qualitatively assessed the extent of tumor-brain adhesion and recorded their observations in the surgical records. The degree of tumor adherence at surgery was divided into three categories: (a) no adhesion: total separation of tumor and brain, which allowed the tumor to be dissected easily from the brainstem and/or cerebellum; (b) partial adhesion: focal zones where the tumor was adherent to the brainstem and/or cerebellum; and (c) complete adhesion to the brainstem and/or cerebellum with no clear plane of dissection around the entire tumor.

MR Imaging–assisted Prediction of Tumor-Brain Adhesion

The MR imaging protocol included axial T2-weighted fast spin-echo, fast imaging employing steady-state acquisition (FIESTA), T2-weighted fluid-attenuated inversion recovery (FLAIR), and T1-weighted spin-echo images acquired after administration of contrast agent. The typical imaging parameters included T2-weighted fast spin echo (repetition time msec/echo time msec, 3000–6000/102; echo train length, 12; 256 × 256 matrix; field of view, 22 cm; and section thickness, 4 mm), FIESTA (7.6/minimum full echo time; 50° flip angle; 256 × 256 matrix; field of view, 12 cm; and section thickness, 1 mm), T2-weighted FLAIR (11 000/147; inversion time, 2250 msec; 256 × 192 matrix; field of view, 22 cm; and section thickness, 4 mm), and T1-weighted spin echo after administration of contrast agent (400–700/minimum full echo time; 256 × 192 matrix; field of view, 22 cm; and section thickness, 4 mm). To assess the capability of traditional MR imaging to predict tumor adherence, T2-weighted fast spin-echo, FIESTA, and T2-weighted FLAIR images were reviewed. A peritumoral CSF cleft was defined as a thin layer of high signal intensity on T2-weighted and/or FIESTA images and low signal intensity on T2-weighted FLAIR images at the tumor-brainstem or tumor-cerebellum interfaces. The presence of these findings, which included the extent of the cleft or absence of the CSF cleft and slip interface, was evaluated by a neuroradiologist (J.H., with 26 years of experience in neuroradiology) who was blinded to the surgical findings. Depending on the extend of the cleft over the tumor surface, it was classified into one of the following three categories: (a) complete cleft: a CSF cleft that could be traced around the whole tumor surface; (b) partial cleft: a CSF cleft that could be traced around 10%–75% of the tumor surface; and (c) no cleft: no CSF cleft could be identified around the tumor surface. The CSF cleft prediction of the degree of adhesion was compared with the surgical findings.

SII Technique

The SII technique is based on MR elastography techniques and uses a single-shot, flow-compensated, spin-echo, echo-planar imaging MR elastography pulse sequence (22,23). SII was performed with a 3-T MR imager (Signa Excite; GE Healthcare, Waukesha, Wis) in three steps. Low-amplitude mechanical vibrations at 60 Hz were generated by an active driver (located outside of the imaging room) and conducted to a soft, pillow-like passive driver placed under the patient’s head in a standard eight-channel receive-only head coil (24) (Fig 1, A). Subsequently, the resulting shear waves produced in the brain were imaged by using the spin-echo echo-planar imaging MR elastography pulse sequence with the motion-encoding gradients synchronized with the applied vibration. The data acquisition parameters were as follows: 3600/62; field of view, 24 cm; bandwidth, 250 kHz; 72 × 72 imaging matrix reconstructed to 80 × 80; parallel imaging acceleration factor of three; and 48 contiguous sections (axial with a 3-mm section thickness). Motion-encoding gradients with amplitude of 40 mT/m, duration of 18.2 msec, and zeroth- and first-order moment nulling were placed on each side of the refocusing radiofrequency pulse and synchronized to the motion. The motion encoding gradients were applied in the positive and negative x-, y-, and z-axis directions with eight equally spaced phase offsets sampled over one period of the 60-Hz motion. The imaging time was less than 7 minutes. The acquired shear motion data were then processed to make visible and quantify the tumor adhesion, which will be described in this article.

Figure 1:

Concept of SII. A, Shear waves are introduced into the brain by using a soft, pillow-like driver placed under the patient’s head, with vibration direction indicated by double-headed arrow. The dotted vertical lines indicate the image plane. B, Assuming a nonadhesive tumor, low intensity shear lines can be made visible because of the IVPD effect. The large differential motion between the tumor and the adjacent brain can create phase variations across the interface, which leads to magnitude signal loss. C, Because of the independent motion of the tumor and the brain, differential motion between the tissues at the interface can be detected, which leads to large OSS.

Image and Data Processing

Two different algorithms (shear line analysis and OSS calculation) were performed to both make visible and quantify the tumor adhesion. The first approach is on the basis of the phenomenon of motion-induced shear lines caused by IVPD (Fig 1, B). The MR signal of a voxel is a vector sum of the magnetization of a large number of isochromats present within that voxel. As illustrated in Figure 1, B, if a low-friction, nonadhesive interface exists, the large differential motion between the tumor and the adjacent brain creates phase variations across the interface, which leads to signal loss within the voxels at the interface. Pseudomagnitude filter analysis was used to make visible these regions, as previously described (21). Briefly, for each of the x-, y-, and z-axis motion-encoding directions, the complex-valued phase-difference images for each phase offset were calculated as the product of the MR images with positive motion encoding and the complex conjugate of the images with negative motion encoding. The resulting images were used to form a new time series of MR images with unit magnitude and phase equal to the phase of the complex images just created. Setting the magnitude to 1 removes any inherent T1-, T2-, or proton density–weighted signal contrast from the original image acquisition so that magnitude signal variations after further processing will be because of local phase variations. To enhance the effect of IVPD, the new complex MR images were then filtered by using a Gaussian low-pass filter (standard deviation of the point-spread function, 1.88 mm) to increase the effective voxel size of the images. The filter size was chosen subjectively while the nine cases were analyzed (while remaining blinded to the surgical findings and the neuroradiologist’s report) to produce filtered images with good shear line contrast and spatial resolution. The mean of the temporally resolved eight filtered magnitude images along the x-, y-, and z-axis directions was calculated independently, and a final shear line image was created from the minimum of the mean of the x-, y-, and z-axis data (21). The presence of a nonadhesive interface is made visible as low intensity lines in the final processed images.

The second approach quantifies the degree of shear displacement discontinuity at the tumor-brain interface by calculating the OSS (20). The OSS is a single value that represents the maximum shear strain occurring at a point across all possible planes (19). The OSS values can thus provide a measure of the amount of change in shear displacement at the tumor-brain interface. An imperfectly bounded interface affects the shear wave propagation (25) and exhibits a shear displacement discontinuity, or slip, across the interface, which is assumed to be related to the stress traction across the surface (26). When a tumor is adherent to the adjacent brain, it is reasonable to assume that the change in displacement is only because of deformation around the tumor-brain interface, and thus would be much smaller than with a loosely bound tumor (Fig 1, C), which would lead to a low OSS value. The OSS was calculated for each phase offset as described in McGarry et al (20), and a final OSS map was calculated from the mean of the eight sets of OSS data. The presence of a nonadhesive interface can be quantified as higher OSS around the tumor interface relative to the adjacent tumor and brain tissues themselves. To assess the OSS values of tumor-brain interfaces, the boundary of the tumor was manually traced on corresponding MR elastography magnitude images. The tumor outlines were then registered to the OSS maps. For each patient, the mean and standard deviations of the OSS values along the outline around the entire tumor surface were reported.

SII Prediction of Tumor-Brain Adhesion

On the basis of information acquired from both proposed methods, the presence of low-intensity shear lines and large OSS values at the interface represents separation of the tumor from the adjacent brain tissue. The degree of tumor-brain adhesion predicted by SII was classified as follows: (a) complete slip interface, in which a dark shear line and higher OSS values can be seen along the whole tumor surface, and indicates no adhesion; (b) partial slip interface, in which a shear line and higher OSS values can be traced between 10% and 75% of the tumor surface; and (c) no slip interface, in which neither shear lines nor contours of visibly higher OSS can be traced on any part of the tumor surface, and indicates complete adhesion. The SII prediction of the degree of adhesion was compared with the surgical findings.

Statistical Analysis

Agreement between the SII prediction and surgical findings and between the CSF cleft prediction and surgical findings were assessed by using Cohen κ coefficients (<0.20, poor agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; and >0.60, good agreement [27]). The pair-wise differences between the prediction of adhesion degree that used CSF cleft and SII were examined by using the McNemar test. P values less than .05 indicated statistical significance. The Wilson score confidence interval (CI) was computed for all proportions calculated in this study. All statistical analyses were performed by using software (JMP version 11.0.0; SAS Institute, Cary, NC).

Results

Pathologic analysis confirmed that all tumors were vestibular schwannomas. The surgical findings and the predictions of tumor adhesion by using SII and peritumoral CSF cleft are summarized in Table 1 for each patient. Intraoperative assessment of tumor adhesion in the nine patients included four cases with no adhesion, three cases with partial adhesion, and two cases with complete adhesion. The range of OSS values along the tumor boundary for each patient is shown in a box plot (Fig 2). We observed a trend that the nonadhesive tumors typically had higher values than the partially and completely adhering tumors.

Table 1.

Summary of the Surgical Findings and the Prediction of Tumor Adhesion by Using SII and the Peritumoral CSF Cleft Sign

Note.—OSS data are means ± standard deviation.

Figure 2:

Boxplot of the OSS values along the tumor boundary for each patient. Boxes represent the 25th and 75th percentiles, with the medians indicated by the middle line in the box. Vertical end bars denote the range of data except for the outliers (ie, values more than 1.5 times of the box length).

As shown in Table 2, the SII predictions about adhesion were concordant with the surgical findings in eight cases (89% [95% CI: 57%, 98%]); five cases had complete slip interface, three cases had partial slip interface, and one case had no slip interface. The presence of the peritumoral CSF cleft sign on T2-weighted, FIESTA, and T2-weighted FLAIR images correlated with the surgical findings in four cases (44% [95% CI: 19%, 73%]); one case was judged as complete cleft, four cases with partial cleft, and four cases with no cleft. The κ coefficient indicates good agreement between the SII predictions and surgical findings (0.82 [95% CI: 0.5, 1]), but only fair agreement between CSF cleft predictions and surgical findings (0.21 [95% CI: −0.21, 0.63]). In this study, however, the difference between the SII prediction and the CSF cleft prediction was not significant (P = .103; McNemar test).

Table 2.

Case Agreement and κ Coefficient for Agreement between Peritumoral CSF Cleft and Surgical Findings and between SII and Surgical Findings

Note.—Data in parentheses are 95% CI range. P value is .103 and was calculated from a comparison of CSF cleft status and SII status by using the McNemar test.

Representative images from SII and conventional MR imaging examinations of vestibular schwannomas are presented in Figures 3–6. As shown in cases 1 and 2 (Figs 3 and 4), low-intensity shear lines and large OSS values were observed at tumor-brain interfaces in patients with no adhesion found at surgery. A complete CSF cleft was observed in case 1 (Fig 3) and a partial CSF cleft was observed in case 2 (Fig 4). On the other hand, in case 8 (Fig 5), the absence of shear lines or contours of visibly higher OSS and no CSF cleft surrounding the tumor agreed with the surgical report of complete tumor adhesion. Discordance of the results occurred in one patient with a heterogeneous tumor (case 9, Fig 6). The presence of shear lines and high OSS values predicted no adhesion, while no CSF cleft was identified and complete adhesion was found at surgery.

Figure 3a:

Images in a 62-year-old woman with a left vestibular schwannoma (case 1). The CSF cleft is observed around the entire tumor-brain interface and seen as (a) high signal intensity on the FIESTA image and (b) low signal intensity on the T2-weighted FLAIR image. This CSF cleft is classified as a complete cleft. The slip interface can be observed as a (c) low-signal shear line with (d) large OSS values around the whole surface of the tumor, indicating independent motion between the tumor and the adjacent tissues, classified as a complete slip interface. Surgical findings demonstrated no adhesion between the tumor-brainstem and tumor-cerebellum interfaces, in agreement with the imaging and SII predictions.

Figure 6a:

Images in a 43-year-old woman with a left vestibular schwannoma (case 9). (a) Axial T2-weighted fast-spin-echo and (b) T2-weighted FLAIR images show no CSF cleft. (c) Axial T1-weighted image after administration of contrast agent shows the heterogeneous solid and cystic mass (arrowhead). (d) Multiple low-intensity shear lines and (e) contours of visibly higher OSS can be seen associated with the tumor, which suggests that there is no adhesion (arrows). However, surgical findings demonstrated extreme adhesion at the tumor-brainstem and tumor-cerebellum interfaces. This mismatch with the SII may reflect internal shearing between intratumoral tissue planes.

Figure 4a:

Images in a 37-year-old woman with a right vestibular schwannoma (case 2). CSF cleft can be seen partially around the tumor surface (arrows) with the absence of a CSF cleft against the brainstem (arrowheads) on (a) FIESTA and (b) T2-weighted FLAIR images. This is classified as a partial cleft. SII shows (c) a low-signal shear line and (d) large OSS values around the whole surface of the tumor, which suggests no adhesion and is classified as a complete slip interface. Surgical findings demonstrated no adhesion at the tumor-brainstem and tumor-cerebellum interfaces, consistent with the SII prediction.

Figure 5a:

Images in a 77-year-old woman with a right vestibular schwannoma (case 8). (a) FIESTA and (b) T2-weighted FLAIR images show no CSF cleft. Peritumoral edema can be observed as high signal intensity on the T2-weighted FLAIR image medially (arrows). Separation between the tumor and cerebellum is seen laterally (arrowheads). No (c) shear lines or (d) large OSS values can be seen around the tumor, which indicates complete adhesion. This case is classified as no slip interface, which correlated with the surgical findings of extreme adhesion at the tumor-brainstem and tumor-cerebellum interfaces.

Figure 3b:

Images in a 62-year-old woman with a left vestibular schwannoma (case 1). The CSF cleft is observed around the entire tumor-brain interface and seen as (a) high signal intensity on the FIESTA image and (b) low signal intensity on the T2-weighted FLAIR image. This CSF cleft is classified as a complete cleft. The slip interface can be observed as a (c) low-signal shear line with (d) large OSS values around the whole surface of the tumor, indicating independent motion between the tumor and the adjacent tissues, classified as a complete slip interface. Surgical findings demonstrated no adhesion between the tumor-brainstem and tumor-cerebellum interfaces, in agreement with the imaging and SII predictions.

Figure 3c:

Images in a 62-year-old woman with a left vestibular schwannoma (case 1). The CSF cleft is observed around the entire tumor-brain interface and seen as (a) high signal intensity on the FIESTA image and (b) low signal intensity on the T2-weighted FLAIR image. This CSF cleft is classified as a complete cleft. The slip interface can be observed as a (c) low-signal shear line with (d) large OSS values around the whole surface of the tumor, indicating independent motion between the tumor and the adjacent tissues, classified as a complete slip interface. Surgical findings demonstrated no adhesion between the tumor-brainstem and tumor-cerebellum interfaces, in agreement with the imaging and SII predictions.

Figure 3d:

Images in a 62-year-old woman with a left vestibular schwannoma (case 1). The CSF cleft is observed around the entire tumor-brain interface and seen as (a) high signal intensity on the FIESTA image and (b) low signal intensity on the T2-weighted FLAIR image. This CSF cleft is classified as a complete cleft. The slip interface can be observed as a (c) low-signal shear line with (d) large OSS values around the whole surface of the tumor, indicating independent motion between the tumor and the adjacent tissues, classified as a complete slip interface. Surgical findings demonstrated no adhesion between the tumor-brainstem and tumor-cerebellum interfaces, in agreement with the imaging and SII predictions.

Figure 4c:

Images in a 37-year-old woman with a right vestibular schwannoma (case 2). CSF cleft can be seen partially around the tumor surface (arrows) with the absence of a CSF cleft against the brainstem (arrowheads) on (a) FIESTA and (b) T2-weighted FLAIR images. This is classified as a partial cleft. SII shows (c) a low-signal shear line and (d) large OSS values around the whole surface of the tumor, which suggests no adhesion and is classified as a complete slip interface. Surgical findings demonstrated no adhesion at the tumor-brainstem and tumor-cerebellum interfaces, consistent with the SII prediction.

Figure 4d:

Images in a 37-year-old woman with a right vestibular schwannoma (case 2). CSF cleft can be seen partially around the tumor surface (arrows) with the absence of a CSF cleft against the brainstem (arrowheads) on (a) FIESTA and (b) T2-weighted FLAIR images. This is classified as a partial cleft. SII shows (c) a low-signal shear line and (d) large OSS values around the whole surface of the tumor, which suggests no adhesion and is classified as a complete slip interface. Surgical findings demonstrated no adhesion at the tumor-brainstem and tumor-cerebellum interfaces, consistent with the SII prediction.

Figure 5b:

Images in a 77-year-old woman with a right vestibular schwannoma (case 8). (a) FIESTA and (b) T2-weighted FLAIR images show no CSF cleft. Peritumoral edema can be observed as high signal intensity on the T2-weighted FLAIR image medially (arrows). Separation between the tumor and cerebellum is seen laterally (arrowheads). No (c) shear lines or (d) large OSS values can be seen around the tumor, which indicates complete adhesion. This case is classified as no slip interface, which correlated with the surgical findings of extreme adhesion at the tumor-brainstem and tumor-cerebellum interfaces.

Figure 5c:

Images in a 77-year-old woman with a right vestibular schwannoma (case 8). (a) FIESTA and (b) T2-weighted FLAIR images show no CSF cleft. Peritumoral edema can be observed as high signal intensity on the T2-weighted FLAIR image medially (arrows). Separation between the tumor and cerebellum is seen laterally (arrowheads). No (c) shear lines or (d) large OSS values can be seen around the tumor, which indicates complete adhesion. This case is classified as no slip interface, which correlated with the surgical findings of extreme adhesion at the tumor-brainstem and tumor-cerebellum interfaces.

Figure 5d:

Images in a 77-year-old woman with a right vestibular schwannoma (case 8). (a) FIESTA and (b) T2-weighted FLAIR images show no CSF cleft. Peritumoral edema can be observed as high signal intensity on the T2-weighted FLAIR image medially (arrows). Separation between the tumor and cerebellum is seen laterally (arrowheads). No (c) shear lines or (d) large OSS values can be seen around the tumor, which indicates complete adhesion. This case is classified as no slip interface, which correlated with the surgical findings of extreme adhesion at the tumor-brainstem and tumor-cerebellum interfaces.

Figure 6b:

Images in a 43-year-old woman with a left vestibular schwannoma (case 9). (a) Axial T2-weighted fast-spin-echo and (b) T2-weighted FLAIR images show no CSF cleft. (c) Axial T1-weighted image after administration of contrast agent shows the heterogeneous solid and cystic mass (arrowhead). (d) Multiple low-intensity shear lines and (e) contours of visibly higher OSS can be seen associated with the tumor, which suggests that there is no adhesion (arrows). However, surgical findings demonstrated extreme adhesion at the tumor-brainstem and tumor-cerebellum interfaces. This mismatch with the SII may reflect internal shearing between intratumoral tissue planes.

Figure 6e:

Images in a 43-year-old woman with a left vestibular schwannoma (case 9). (a) Axial T2-weighted fast-spin-echo and (b) T2-weighted FLAIR images show no CSF cleft. (c) Axial T1-weighted image after administration of contrast agent shows the heterogeneous solid and cystic mass (arrowhead). (d) Multiple low-intensity shear lines and (e) contours of visibly higher OSS can be seen associated with the tumor, which suggests that there is no adhesion (arrows). However, surgical findings demonstrated extreme adhesion at the tumor-brainstem and tumor-cerebellum interfaces. This mismatch with the SII may reflect internal shearing between intratumoral tissue planes.

Discussion

Extra-axial intracranial tumors, such as vestibular shwannomas and meningiomas, impose surgical risks in a variety of ways. In general, with increasing posterior fossa tumor size and the associated mass effect, the risk of surgery increases. Further affecting risk is the absence of an arachnoid plane and resulting adhesion between a tumor and the brain, brainstem, or adjacent cranial nerves. When this arachnoid plane is absent, the tumor often develops a pial blood supply. Consequently, the dissection of the tumor from adjacent structures is tedious and substantially increases the risk of stroke or damage to the brainstem and cranial nerves. To understand the degree of adherence preoperatively would allow the surgeon to more accurately inform the patient of his or her risk for a permanent deficit. For instance, patients with adherent tumors could be advised of a prolonged surgical procedure, increased risk of surgical complication, and a greater likelihood of less than gross total resection. Knowledge of tumor adhesion might change clinical management (eg, posterior fossa tumors less than 2.5 cm in diameter where SII indicates an adherent tumor could be directed to stereotactic radiosurgery instead of microsurgery).

Conventional MR imaging approaches to predict the adhesion between the tumor and the brain rely on depiction of the gross morphologic structure of the tumor-brain interface and the presence or absence of a CSF cleft around the interface (8,11–13). The CSF cleft occurs when CSF fills in the spaces between the tumor and the adjacent brain. It is typically observed as a thin layer of high signal intensity on T2-weighted and FIESTA images and low signal intensity on T2-weighted FLAIR images. Although the MR signal intensity characteristics of the CSF cleft can provide information about tumor adherence, the absence of CSF cleft at the interface is not sufficient to predict adhesion. For example, Figures 4a and 4b show a nonadhesive tumor that exhibits no CSF cleft on a segment of the tumor-brainstem interface. This suggests that standard MR images alone that use T2-weighted, FIESTA, and T2-weighted FLAIR cannot reliably predict the degree and locations of tumor adhesion, especially in the case of partial adhesion (one of three was correctly predicted as partial adhesion in this study).

Figure 4b:

Images in a 37-year-old woman with a right vestibular schwannoma (case 2). CSF cleft can be seen partially around the tumor surface (arrows) with the absence of a CSF cleft against the brainstem (arrowheads) on (a) FIESTA and (b) T2-weighted FLAIR images. This is classified as a partial cleft. SII shows (c) a low-signal shear line and (d) large OSS values around the whole surface of the tumor, which suggests no adhesion and is classified as a complete slip interface. Surgical findings demonstrated no adhesion at the tumor-brainstem and tumor-cerebellum interfaces, consistent with the SII prediction.

Compared with conventional MR imaging, which is based on the static characteristics of tissue, the dynamic SII technique presented in this work provides a direct measure of adhesion by assessing the shear connectivity across the interface between a tumor and the adjacent brain tissue. An adhesive interface can exhibit shear displacement continuity across the interface (ie, the tumor and brain tissue on either side of the interface move together when experiencing shear motion). Conversely, a nonadhesive interface can create large differential motion between the tumor and brain because of low friction or a slippery boundary. The application of SII therefore requires two critical steps: introduction of shear motion into the brain and encoding that motion into the MR signal, which are the two basic steps of MR elastography (22,23). We therefore refer to SII as an MR elastography-based method. It was shown (21,28) in MR elastography experiments that phase variations within a voxel can produce intravoxel signal attenuation because of IVPD. The presence of a slip interface would increase the amount of differential motion between the tumor and the brain, therefore increasing the amount of phase dispersion and signal attenuation at the interface. A previous study performed on the abdomen and forearm demonstrated the presence of shear interfaces with the IVPD effect by visualizing peritoneal bowel interfaces and functional compartment interfaces of the forearm musculature, respectively (18). While the shear line image provides an indicator of slip interfaces, the OSS map can quantify the degree of adhesion. Because the OSS reflects the maximum shear strain that occurs in the plane on which the maximum shear stress is reached, the large OSS values at the interface correspond to large shear deformation across the adjacent tissue layers, which suggest a slippery and nonadhesive interface.

In our study, the prediction of brain and vestibular schwannoma adhesion by SII versus conventional MR images agreed with the surgical findings in, respectively, eight of nine patients (κ = 0.83) versus four of nine patients (κ = 0.21). A higher κ coefficient indicates a better agreement between SII predictions and surgical findings. However, when the McNemar test was used to compare the difference between the SII and CSF cleft predictions, statistical significance was not reached (P = .103), possibly because of the small number of patients (n = 9). In future studies, the sensitivity, specificity, and accuracy for prediction of degree of tumor adhesion by using SII and CSF cleft will be tested with a larger patient group.

Although our results indicate that SII is a reliable method to predict tumor adhesion, one of the nine patients was discordant. With case 9, both the shear line image and the OSS map suggested separation of the tumor from the adjacent brainstem and cerebellum. However, the surgeon found complete adherence of this tumor. The discrepancy was likely because of the complicated internal tissue motion that was from the heterogeneity of the tumor, which is evident from the multiple internal cysts and tissue planes (Fig 6a–6c). Because the contrast of SII is derived from the differential shear wave motion of tissues on either side of an interface, the existence of inhomogeneities inside the tumor likely resulted in complicated shearing between cystic and solid intratumoral tissue planes. These regions correspond to the areas in the shear line image with multiple shear lines (Fig 6d). In the future, optimization of the choice of vibration amplitude and frequency and imaging parameters to further increase the sensitivity of SII may better depict internal structures and tumor-brain boundaries of these lesions.

Figure 6c:

Images in a 43-year-old woman with a left vestibular schwannoma (case 9). (a) Axial T2-weighted fast-spin-echo and (b) T2-weighted FLAIR images show no CSF cleft. (c) Axial T1-weighted image after administration of contrast agent shows the heterogeneous solid and cystic mass (arrowhead). (d) Multiple low-intensity shear lines and (e) contours of visibly higher OSS can be seen associated with the tumor, which suggests that there is no adhesion (arrows). However, surgical findings demonstrated extreme adhesion at the tumor-brainstem and tumor-cerebellum interfaces. This mismatch with the SII may reflect internal shearing between intratumoral tissue planes.

Figure 6d:

Images in a 43-year-old woman with a left vestibular schwannoma (case 9). (a) Axial T2-weighted fast-spin-echo and (b) T2-weighted FLAIR images show no CSF cleft. (c) Axial T1-weighted image after administration of contrast agent shows the heterogeneous solid and cystic mass (arrowhead). (d) Multiple low-intensity shear lines and (e) contours of visibly higher OSS can be seen associated with the tumor, which suggests that there is no adhesion (arrows). However, surgical findings demonstrated extreme adhesion at the tumor-brainstem and tumor-cerebellum interfaces. This mismatch with the SII may reflect internal shearing between intratumoral tissue planes.

It should be also noted that there is some amount of IVPD everywhere in the brain in an MR elastography acquisition, even within homogeneous tissue, because the propagating shear waves may be at different phases within a voxel, which causes displacements to differ from one part of the voxel to another. This would cause a spread of phase in the MR signal and thus a slightly lower magnitude than if the tissue were static. Heterogeneity does not necessarily increase or decrease this effect. However, at a slip interface, the displacements on either side of the boundary are much larger, which causes greatly increased IVPD. The same observations apply to the OSS measure.

Another technique was developed to characterize tissue interfaces. Taoka et al (15,16) described an MR imaging-based technique for assessing meningioma-brain adhesion by subtracting two MR images obtained in different phases of the cardiac cycle (17). The cardiac cycle–dependent motion amplitude is variable in different intracranial locations; this is a limitation of that technique. The approach tested here, however, uses extrinsically applied mechanical vibrations at a much higher frequency and a measurement approach that is more sensitive to micron-range shear motions.

In SII, shear line imaging and OSS mapping can be complementary approaches for the visualization and prediction of tumor adhesion. Because shear line imaging does not require phase data to be unwrapped, it requires minimal data processing time. The contrast of shear line imaging depends on both the amount of phase variation across the interface and the size of the chosen filter. A Gaussian low-pass filter with a point-spread function of 1.88 mm (standard deviation) was used in the our study to increase the shear line contrast by increasing the effective voxel size of the interface, which allowed more phase variations within the voxels (21). The choice of filter size is a trade-off between spatial resolution and good shear line contrast. The filter size must be adjusted to produce a significant amount of shear line contrast while maintaining reasonable image resolution. OSS provides a quantitative technique to depict the degree of adhesion. The calculation was performed directly on the displacement data and does not require a low-pass filter. This maintains resolution close to the original data but does require unwrapped phase data, which increases the processing time.

In this study, eight phase offsets of x, y, and z positively and negatively motion-encoded data were performed (29). The shear line images and OSS maps were calculated from the average of the eight offsets of low-pass–filtered images. However, SII may be accelerated by acquiring fewer phase offsets, such as only two time offsets separated in time by a quarter of a period of the motion, which could reduce the scan time by a factor of four. Moreover, the OSS calculation in the study depends on the amplitude of the shear motion. Although we applied the same vibration frequency to each patient and tried to deliver the same amplitude of motion to each patient, there may still be amplitude variations within the brain because of wave attenuation and scattering and between patients because of differences in how the patient’s head was placed on the driver. Future studies that use OSS for SII will focus on development of a reliable wave amplitude normalization method to reduce these types of variations.

As discussed, the contrast of shear line images is dependent on the amount of IVPD induced (ie, the local phase gradient across the interface). As the phase gradient increases the IVPD increases and more signal loss occurs. The phase gradient is related to the local shear wavelength and amplitude of motion such that high-amplitude motion and short wavelengths lead to a large phase gradient and increased IVPD in shear line images (21,28). At higher frequencies of motion, shear wavelengths are smaller. Therefore, higher vibration amplitude and frequency are more desirable to increase IVPD. However, there are some limitations to the upper amplitude and frequency that can be used in human brain MR elastography examinations. Typical vibration amplitude within the safety range is on the order of tens of microns (30) and the usable frequency range is about 25–100 Hz with higher attenuation of shear waves occurring at higher frequencies. Future developments that use multiple drivers may partly compensate for wave attenuation at higher frequencies by producing more uniform wave amplitude through the head (31) or the opportunity to create unique shear wave polarizations and propagation directions at lower frequencies that will increase the amount of local tissue shearing at tumor boundaries.

This study had several limitations. First, the reference standard of this study is the qualitative assessment of tumor adherence by the surgeons during resection. In future studies with larger numbers of tumors, a quantitative scale on the basis of degree and the location of tumor adherence will be incorporated into the tumor resection operative report. Second, data acquisitions were limited by relatively low resolution. A higher resolution acquisition with higher vibration frequency and more uniform vibration from multiple motion drivers may allow for more accurate measurements in heterogeneous tumors. Ideally, the technique would be able to predict adhesion between the tumor and the adjacent cranial nerves and vessels, and the brain tissue itself.

SII, an MR elastography-based technique, provides a method to preoperatively determine the degree of tumor-brain adhesion in patients with vestibular schwannomas that offers a potential approach to improve preoperative planning, including determination of surgical risk and likelihood of gross total resection. While these results are encouraging, further studies are needed in a larger population to confirm the accuracy of SII predictions for tumor-brain adhesion.

Advances in Knowledge

• ■ Slip interface imaging (SII), a newly developed MR elastography–based imaging technique, provides a method for preoperatively determining the degree of tumor-brain adhesion.

• ■ In vestibular schwannomas, SII predictions were confirmed at surgery in eight of nine patients (good agreement with the κ coefficient = 0.82, 95% confidence interval: 0.5, 1).

Implication for Patient Care

• ■ SII may provide a method for more accurate preoperative prediction of the completeness of tumor resection and determination of surgical risk in patients with vestibular schwannomas.