Readout-segmented echo-planar diffusion-weighted imaging improves geometric performance for image-guided radiation therapy of pelvic tumors

Abstract

Background and purpose: Diffusion-weighted imaging using echo-planar imaging (EPI) is prone to geometric inaccuracy, which may limit application to image-guided radiation therapy planning, as well as for voxel-based quantitative multi-parametric or multi-modal approaches. This research investigates pelvic applications at 3 T of a standard single-shot (ssEPI) and a prototype readout-segmented (rsEPI) technique.

Materials and methods: Apparent diffusion coefficient (ADC) accuracy and geometric performance of rsEPI and ssEPI were compared using phantoms, and in vivo, involving 8 patients prior to MR-guided brachytherapy for locally advanced cervical cancer, and 19 patients with prostate cancer planned for tumor-targeted radiotherapy. Global and local deviations in geometric performance were tested using Dice Similarity Coefficients (DC) and Hausdorff Distances (HD).

Results: In cervix patients, DC increased from 0.76±0.14 to 0.91±0.05 for the high risk clinical target volume, and 0.62±0.26 to 0.85±0.08 for the gross tumor target volume. Tumors in the peripheral zone of the prostate gland were partly projected erroneously outside of the posterior anatomic boundary of the gland by 3.1±1.6 mm in 11 of 19 patients using ADC-ssEPI but not with ADC-rsEPI.

Conclusions: Both cervix and prostate ssEPI are prone to clinically relevant geometric distortions at 3T. rsEPI provides improved geometric performance without post-processing.

Keywords: Cervical cancer; Diffusion weighted imaging; Geometric distortion; Image-guided radiotherapy; MRI; Prostate cancer.