49 Performance Assessment of Magnetic Resonance Imaging (MRI) pelvimetry using ex vivo anatomic pieces and a 3D-printed geometric-distortion phantom

Introduction Feto-maternal morbidity and mortality increase with prolonged labor which can be related with fetalpelvic disproportion. Pelvimetric imaging is useful for assessing the eligibility of vaginal delivery in patients suspected of having disproportion [1] . Commonly performed using Computed Tomography (CT), pelvimetry is being increasingly done in MRI in order to tackle issues of ionizing-radiation exposure and take advantage of the rich contrast provided in MRI [2] . Nonetheless, MR images can be subjected to geometric distortions because of hardware design limitations [3] . Such distortions may impair MRIpelvimetric measurements. In this study, we assessed the accuracy of MRI-pelvimetry compared with CTpelvimetry using ex-vivo pelvises and a 3D-printed geometric-distortion phantom. Method We scanned ten ex-vivo pelvises on a CT scanner (Siemens Sensation) using the implemented clinical low-dose (LD-CT) and high-dose (HD-CT) pelvimetry protocol. MRI was performed on four 1.5T MRI scanners (Siemens Avanto, GE Signa, Philips Achieva, Toshiba Elan). Pelvises were scanned in sagittal and oblique axial using T1-, T2-weighted FSE and T1-weighted SPGR sequences. A trained radiologist performed pelvimetry measurements. Pelvic inlet and outlet anteroposterior diameters were measured using sagittal images. Pelvic inlet and outlet transverse diameters were measured using oblique axial images. For data analysis, MRI measurements were compared with CT using Bland-Altman plots and discrepancy analysis. MRI measurements were also analyzed for inter-scanner variability. Geometric distortion (GD) measurement: we used a custom previously-published methodology which relies on a 3D-printed phantom made up of 3621 control points which are 5 mm-diameter spheres equally spaced within a cylinder (diameter = length = 170 mm) [4] . GD images were acquired on the Siemens Avanto MR unit with and without the manufacturer’s correction that is available but commonly unused by clinical teams. Results The discrepancy between LD-CT and HD-CT measurements was up to 10%. The discrepancy between MRI and LD-CT was up to 11%. As for MRI and HD-CT, the discrepancy was up to 7%. MRI inter-scanner variation was up to 8%. As for geometric distortions, measurements were −1.4 mm  δ δ  0.3 mm with correction. Conclusion For pelvimetry purposes, our preliminary results show that MRI is as much accurate as high-dose CT and outperforms low-dose CT. We also show that geometric distortions can be minimized in MR images by using the implemented corrections. Hence MRI pelvimetry appears as an excellent alternative to CT pelvimetry in an endeavor to reduce fetuses exposure to ionizing radiation. Upcoming works in this study include assessment of interobserver reproducibility with the intervention of a second radiologist and geometric distortion measurement on the three remaining MRI scanners.