Session Item

Needle localization in MRI-guided brachytherapy is challenging due to various uncertainties, including lack of signal on standard MRI. Therefore, in the clinical setup CT images are utilized for needle localization, while MR images are used for target and organs at risk delineation. In this phantom study, we used a Pointwise Encoding Time Reduction with Radial Acquisition (PETRA) sequence for needle detection using MRI only. Accuracy of needle localization was assessed by comparison to gold standard CT images.

A gelatin-based phantom with 10 parallel plastic ProGuide Sharp Needles 6Fx294mm (Elekta Brachytherapy, Netherlands) was imaged in a 3T SIEMENS Vida simulator using a spine matrix array coil topped with a Qfix Insight board, and a Body 18 long coil. The phantom consisted of a plexiglass box featuring an orientation verification marker and two identical templates with 2mm diameter holes. Nine parallel empty needles and one with a metallic trocar for metal artifact observation were placed in the templates. Isotropic 0.82 mm edge length voxels were acquired with a coronal 3D PETRA (TR\TE\TI  3.32\0.07\100 ms, FOV 289*289 mm², BW 400 Hz/px, 100000 radial views). 2mm thick coronal T2-weighted Turbo Spin Echo (TSE) slices (TR\TE 3840\102 ms, FOV 199*199 mm², matrix 736*736, BW 200 Hz/px) were acquired for comparison. The phantom with needles was subsequently scanned on a helical CT (250 mA, 120 kV, 1.25 mm slice thickness). Needles were localized in the Oncentra Brachy (Elekta Brachytherapy, Netherlands) treatment planning system using both CT and MRI. CT images were registered to PETRA images, and needle positions were evaluated.

Plastic needles showed positive signal on PETRA (acquisition time 6min 20s) and less distortions than on conventional TSE images (example comparison at a similar slice location on picture below). The metal artifact extended symmetrically around the trocar on PETRA, contrary to its contorted shape on TSE.

Needle localization using PETRA MRI only was uncomplicated (3D representation using Maximum Intensity Projection (MIP) in a) below). The observed position difference in PETRA and CT images was < 1 mm for the needle tip in all 10 needles, and < 0.5° for the deflection angle, making this level of accuracy clinically acceptable. CT to MR registration (example registered slice showing good agreement of needle positions in b) below) did not reveal any deformation of empty needles on PETRA, suggesting that these MRI images can be used for treatment planning.

We demonstrated that plastic needles used in gynecological brachytherapy can be imaged with submillimeter accuracy using a manufacturer provided PETRA sequence in acceptable scanning time. Future clinical workflow would include registration of PETRA with MR sequences of the same session utilized for segmentation. The results of this work reveal the potential to ultimately replace CT for needle localization and treatment planning in MRI-guided brachytherapy.