Prospective Study of Slip-interface Imaging in Meningioma for Brain-tumor Adhesion

Introduction: Magnetic Resonance Elastography(MRE) is a MRI-based imaging sequence utilizing mechanical waves to quantify the stiffness of tissue measured in kilopascal. More recently, an MRE form has been developed called slip-interface imaging(SII) to evaluate the plane between extra-axial brain tumors and adjacent neural structures. We report on SII in meningioma for tumor-brain adhesion.

Methods: 18 tumors were imaged in 17 patients[mean age 59.7 ± 11.7(38–80) years;10(59%) males]. One tumor was a dural metastasis. MRE data were collected with a spin-echo EPI-MRE pulse sequence on a 3T MR scanner. Shear waves at 60Hz were introduced with a soft pillow-like driver placed under the head. A board-certified neuro-radiologist independently reviewed the SII images for adhesion. Octahedral shear-strain(OSS), a quantitative SII method, was used to determine the tumor-brain interface. On the OSS map a tumor is considered adherent if there is a low shear deformation value at the tumor-brain interface, while no adhesion results in a high shear value. Experienced meningioma surgeons documented tumor adhesion. Intraoperative observations regarding adherence were graded as present or absent based on whether the tumor folded away from the brain without any dissection. Tumors with separate sections of both adherence and no adherence at surgery and on SII were considered mixed. The anatomical locations of these areas were recorded on MRI images by surgeons and compared with the radiologist interpretation. Statistics included Cohen κ coefficients(0.20, poor agreement;0.21–0.40, fair agreement;0.41–0.60, moderate agreement;>0.60, good agreement), sensitivity, specificity, and positive and negative predicative values(PPV and NPV).

Results: Mean maximum tumor diameter was 4.7 ± 1.1(2.7–6.8)cm. Locations were cerebellopontine angle(n = 1), foramen magnum(n = 1), lateral sphenoid wing(n = 1), medial sphenoid wing(n = 5), olfactory groove(n = 1), parasagittal(n = 8), and temporal convexity(n = 1).

Regarding overall adherence, surgeons found 6(33%) tumors to have no, 8(44%) mixed, and 4(28%) complete adhesion. SII found 7(39%) tumors to have no, 7(39%) mixed, and 4(28%) complete adherence(κ=0.66,p = 0.002). Four tumors were incorrectly categorized as follows: no adhesion at surgery, complete adhesion on SII(n = 1); mixed adhesion at surgery, no adhesion on SII(n = 2); compete adhesion at surgery, mixed adhesion on SII(n = 1). Of the mixed tumors, surgeons found 6 tumors had 2 locations and 2 had 3 locations of different adhesion and SII found 4 tumors had 2 locations and 3 tumors had 3 locations of different adhesion. There were 30 total brain-tumor interfaces for analysis. SII correctly categorized 13 as adherent and 7 as non-adherent(κ=0.31, p = 0.09). For adherence, sensitivity, specificity, positive predictive value, and negative predictive value were 54%, 77%, 63%, and 68% respectively.

In tumors that were discordant, two had significant amounts of edema which can lead to a false positive for no adhesion; one had poor wave images which does not generate enough movement at the brain-tumor interface for OSS calculation; and one had a small area of adhesion at surgery not seen on SII, but the rest of the tumor was categorized correctly.

Conclusion: SII correctly categorized the brain-tumor interface in 78% of tumors overall with good agreement and 66% of all interfaces with fair agreement. With further refinement, SII is a promising technology for determining adhesion at the brain-tumor interface.