Lessons learnt from the first radical cervix treatment on the MR-Linac

Abstract

Purpose or Objective The MR-linac is ideal for cervical cancer radiotherapy due to excellent soft tissue visualisation and the ability to do daily adaptation of the treatment plan, as target position can vary due to bladder, rectal and bowel filling. Occasionally the target can be ante- or retroverted; in these instances when treating on conventional linacs with CBCT, high-risk disease is prioritised for matching, resulting in sub-optimal coverage of the uterus and nodal chains. On the MR-Linac however, targets and OARs are recontoured daily to create a new treatment plan that meets planning objectives. Up to now, only palliative cervical cancers have been treated on the MR-Linac. Here we present the lessons learnt from the first radical cervix treatment. Materials and Methods Patient selection was restricted due to the limited sup-inf field length (22cm), but a suitable node-negative patient was identified. The prescription, contours and reference plan were created as per EMBRACE II guidelines. The primary RTP scan was MR and bulk density overrides were applied to enable dose calculation. Daily adaptation was carried out via the Adapt-to-Shape (ATS) workflow; deformed structures were reviewed and edited prior to a full plan reoptimisation. Results Contouring was initially time consuming, in part due to the 1mm online MR slice thickness, but improved with experience, with treatment sessions eventually concluding in under an hour – see Figure 1. Deformation was acceptable but could perhaps be improved if online MR and RTP slice thickness matched. Only gross changes were made; edits to fix small imperfections were omitted, as these became irrelevant post target margin expansion. However, bladder and bowel bag contours were rigidly propagated. A drinking protocol was utilised so the bladder would fill during the planning process and eventually match or slightly exceed the rigid contour, without significantly affecting the target position. Optimisation times were long initially; in trying to achieve EMBRACE II style sparing whilst maintaining good coverage, hotspots were inadvertently created in the target. On two fractions, the TPS froze during segmentation, requiring termination of the session and restarting the process. Consequently, a new reference plan with more robust optimisation parameters was created (including increasing max number of segments), which reduced optimisation time and improved the dose distribution. Conclusion Lessons were learnt during the first radical cervix treatment on the MR-Linac to improve future treatments. Having a robust reference plan is imperative when undergoing online reoptimisation. Contour deformation was adequate but could be improved by changing slice thickness. To limit contouring times, edits should only be made if clinically significant. The bladder fills during the planning process, therefore rigid propagation of the bladder contour is recommended and drinking protocols can be designed to ensure the bladder is full when treatment starts.