TH‐A‐BRF‐10: Quantification of Subject‐Specific Susceptibility‐Related Geometric Distortion in Clinical Liver MRI for Radiation Therapy

Abstract

Purpose:Subject‐specific susceptibility‐induced B0 inhomogeneity affects the geometric accuracy of MRI images, with potential impact on the accuracy of treatment planning and image guidance. This study quantifies individual distortions for patients with intrahepatic tumors.Methods:Liver MRI scans of 13 patients who were enrolled in an IRB approved study were acquired on a 3T scanner (Skyra, Siemens). Clinical contrast‐enhanced images were acquired by a 3D Volume Interpolated Breath‐Hold Examination (VIBE) sequence with TE/TR =1.65/4.3ms, voxel size = 2.5×2.5×3mm (axial sets) or 3.5×3.5×5mm (coronal sets), and frequency‐encoding (FE) bandwidth of 440Hz/pixel. B0 inhomogeneity was evaluated by acquisition of dual gradient echoes (GRE) with TE1/TE2/TR =4.92/7.38/106ms, a voxel size of 3.5×3.5×3.75mm, a FE bandwidth of 290Hz/pixel and breath‐hold. The phase difference maps from the two echoes were unwrapped using a quantitative MRI analysis software package (FSL, FMRIB, Oxford, UK). The resulting calculated B0 inhomogeneity maps (ΔB0 in Hz) were converted to voxel distortion maps in the FE direction (ΔX in mm) corresponding to the VIBE images.Results:Maximum susceptibility‐induced distortions were observed in the liver dome near the diaphragm. Using results from coronal VIBE images in this study as examples, we observed clusters of voxels displaced close to or greater than ΔX=3.5mm (440 Hz in ΔB0) in the VIBE images of the liver in 3 patients, and greater than 2 mm but less than 3.5 mm in 12 patients. On average, approximately 14% of imaged liver voxels had distortions ΔX>±1mm (ΔB0>±125Hz) and 1% of the voxels had ΔX>±2mm (ΔB0>±250Hz).Conclusion:Although advanced MRI techniques, like VIBE, permit faster acquisitions allowing for breath‐held examinations with limited motion‐induced artifacts, this study suggests that localized distortions due to subject‐specific susceptibility variations can occur and require additional advanced corrections based on measuring patient‐specific B0 inhomogeneity maps.The authors would like to acknowledge funding support from NIH R01EB016079 and NIH/NCI RO1 CA132834.