Technical Note: Feasibility of MRI-based estimation of water-equivalent path length to detect changes in proton range during treatment courses.

Abstract

To evaluate the feasibility of using on-treatment magnetic resonance imaging (MRI) to detect proton range changes during treatment courses. MRI-based virtual computed tomography (vCT) was generated to calculate water-equivalent path length (WEPL) at the distal surface of a clinical target volume. T2-weighted MR images with and without fat suppression were processed by thresholding and fuzzy c-mean clustering to assign a bulk HU (Hounsfield Unit) to each segment of fat, water, and air. Bony tissues in vCT were transferred from planning CT via a region-based registration. We validated this method using images from three patients (aged 9, 12, and 17 yr old) with pelvic sarcomas who underwent proton therapy. MRI-based WEPL was evaluated against those of repeat CT (rCT) and cone beam CT (CBCT). The vCT agreed well with the rCT, wherein the use of bulk HU and imperfect bone registration contributed to discrepancies. The 99th percentile error of the MRI-based WEPL was up to 3.3 mm for posterior fields when we compensated the effect of mobile air. The gamma index (2-mm WEPL/2-mm pixel distance) was greater than 99% for those fields. We observed larger errors for anterior fields, which were due to bowel gas movement, mismatched respiratory motion, and differences in patient posture between the vCT and rCT. Applied to multiple on-treatment MRI of a patient, the estimated WEPL demonstrated an underrange trend, which was consistent with the CBCT results and the increased patient body circumference. The MRI-based vCT produced highly accurate WEPL estimates, demonstrating the potential of using on-treatment MRI for detecting changes in proton ranges.