Session Item

Online adaptive MR-guided radiotherapy is resource intensive, requiring a team of radiation therapists (RTs), medical physicists (MPs), and radiation oncologists (ROs) for each fraction. Implementation of a new treatment paradigm offers the opportunity to evaluate roles for each member of the multidisciplinary team, and optimize the use of each skillset to maximize efficiency. Here we report a three phase practice-based training strategy developed to transition from RO-driven contouring to RT-driven contouring for whole gland prostate MR-Linac radiotherapy to maximize resource efficiency.

In Phase One, seven MR-Linac RTs independently contoured the target and organs-at-risk on T2-weighted MR images from 11 previously treated MR-Linac prostate patients. The case mix was chosen to ensure a broad representation of differing patient anatomy. The Dice similarity coefficient (DSC) was calculated against contours RO generated during the online session. The RO also performed a qualitative contour review of the RT-generated contours using a 5 point Likert scale; a score of 4 or 5 was deemed a pass, and a 90% pass rate was required for Phase One. Phase Two involved RO supervised online contouring, plan generation and assessment during adaptive treatment fractions, with participants requiring a score of autonomy (no direction required by RO) on 10 cases (minimum 8 patients) to advance. The final phase of training involved independent RT-driven contouring and planning, supported by offline contour and plan review by RO prior to the next fraction.

In Phase One, the mean DSC was 0.916 (range 0.847-0.966) for prostate, 0.895 (range 0.645-0.99) for bladder, and 0.981 (range 0.962-0.998) for rectum. Mean Likert scores for the prostate was 3.7 (range 3-4), the bladder was 4.1 (range 3.7-4.6), and the rectum was 4.3 (range 3.6-4.7). Qualitative prostate contour differences included under-delineation at the base, and variation at the apex. Five RTs did not attain a pass rate of 90%, attended follow-up one-on-one review, and subsequently performed additional contours on a further training set of cases (n=5). Each participant completed a median of 12 (range 10 – 13) cases in Phase Two. Minor direction were required from the RO on 5 cases related to target contouring (contour shape, and contour variability at the rectum prostate interface, prostate base, and prostate apex). ROs reviewed 179 treatment fractions in Phase Three over 5 months; 5 cases were marked acceptable but with suggestions for next fraction (4 involved contours, 1 the adapted plan), all other cases were deemed acceptable.

A training strategy was developed to enable a RT-driven workflow for prostate adaptive radiotherapy for MR-Linac treatment. This model may be adapted for other anatomical sites to maximize efficiencies for MR-Linac radiotherapy.