Optimized MR Simulation Protocol for Framed Stereotactic Radiosurgery at 3T

Abstract

Magnetic resonance imaging is often the primary imaging modality for framed stereotactic radiosurgery (SRS). However, eddy currents, chemical shift, gradient nonlinearities, and the magnetic susceptibility distribution of the patient can all contribute to geometric distortions that may affect the accuracy of SRS if not properly controlled. The goal of this work was to establish an optimized MR simulation protocol for framed SRS at 3T, and assess the severity of any residual geometric distortions. Imaging was performed on a 3T scanner using an SRS phantom (CIRS, model 603A). A titanium Leksell head frame was secured to the phantom. The fiducial box was then affixed to the head frame. The combined, phantom, head frame, and fiducial box assembly were then immobilized in treatment position using a holder of our design. The combination of two flexible phase-array coils and the integrated spine RF coil was used for signal reception. Magnetic field mapping was performed using a dual echo GRE sequence. Different shimming modes (tune up, standard, standard + local shim, standard + brain shim, advanced, advanced + local shim, advanced + brain shim) were applied prior to acquisition of each field map. In addition, 3D T1 FLASH images were acquired with different gradient slew rate modes (whisper, normal, fast) and readout bandwidths. Vendor-provided 3D gradient nonlinearity distortion correction was applied to all images prior to them being sent to a commercial deformable registration algorithm system for analysis. Large magnetic field variations were observed with tune up and standard shim modes. The advanced + brain shim mode resulted in smallest magnetic field variations over the brain. Shimming over the brain did not affect the geometric accuracy of the fiducial markers. The low slew rate whisper gradient mode reduced eddy current-induced distortion by 1.5mm compared to normal gradient mode. Reducing the readout bandwidth from 980 Hz/pixel to 440 Hz/pixel reduced eddy current-induced distortion further by 2mm and increased SNR, while minimizing the severity of chemical shift. Eddy currents induced by the gradients in the titanium frame can result in geometric distortions of fiducial that exceed 1mm in regions of the fiducial box within 2cm of the SRS frame. We established an optimized MR simulation protocol for framed SRS at 3T. Our recommendations include: 1) use of low slew rate gradient modes to reduce eddy current induced distortions, 2) reduce readout bandwidths to 440 Hz/pixel as a tradeoff between eddy current induced distortions and water-fat shift, 3) scan with the prescription volume centered at isocenter, 4) perform advanced high-order shimming with shim volume localized to brain, 5) confirm A-P phase encode direction to avoid flow artifacts, 6) apply 3D gradient nonlinearity distortion correction prior to using images for treatment planning, 7) avoid registering images in treatment planning system using fiducials within 2cm of the SRS frame.