(Tuesday, 7/31/2018) 3:45 PM - 4:15 PM
Room: Exhibit Hall | Forum 2
Purpose: one drawback of using MRI as the sole imaging modality in stereotactic radiosurgery (SRS) treatment planning is geometric distortion; MRI images can be compromised by both system- and patient-specific distortions. We investigated the effects of geometric correction of MRI images used for SRS planning on target volume delineation, center of mass shift, and dose coverage.
Methods: We evaluated SNR, signal uniformity and geometric distortion using a 1.5T Siemens scanner with a Tx/Rx RF head coil and Quasar GRID3D MRI phantom. A system distortion map was derived using Quasar GRID 3D software and in-house MATLAB code. Patient-specific correction involved calculating the complex phase difference map with receiver bandwidth and isotropic resolution via a field map technique using a multi-echo gradient echo sequence. Eight patients were MRI- and CT-scanned for SRS planning under an IRB-approved protocol using a 3D post-contrast axial MPRAGE pulse sequence.
Results: In the phantom study our correction algorithm showed improvement in the frequency encoding direction; maximum deviation shrank from 0.68 to 0.41 mm. The target volume of corrected MRI images averaged 11.3%, smaller than non-corrected images, with a CI of 49.2% and 0.45 mm distortion. We observed an average 0.04 mm change in center of mass shift between contours, with maximum shift in the Z-direction (inside the magnet). Twelve targets were contoured for the first patient and the original plan transferred to corrected MRI images for comparison; we found 0.31% 16 Gy under-dose using uncorrected MRI images, but 2.63% overdose for 20 Gy coverage.
Conclusion: Contours based on corrected MRI images differ from those uncorrected images, offering opportunities to evaluate differences in SRS tumor control probability (TCP) and normal tissue complications. Future studies will include more patients; account for inter- and intra-observer variation in contouring; and correlate computed results with clinical data and possibly pathological specimens.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
