Radiographer-Led, CBCT-Guided Online Adaptive Radiotherapy for Muscle Invasive Bladder Cancer Using Artificial Intelligence

Abstract

<h3>Purpose/Objective(s)</h3> The advent of commercially-available CBCT-guided online adaptive radiotherapy (oART) has made daily replanning possible in our center. We investigated the dosimetric impact and clinical feasibility of a treatment radiographer (RTT)-led oART with daily plan re-optimization service for muscle invasive bladder cancer (MIBC) at our center. <h3>Materials/Methods</h3> Between January 2021 and February 2022 ten patients with MIBC were treated with oART. The patients underwent a radiotherapy (RT) planning CT scan with an empty bladder. RT was prescribed to a dose of 55 Gy in 20 daily fractions to the planning target volume (PTV) and delivered using either 7- or 9-field IMRT. CTV to PTV margins were 1.5cm superiorly/anteriorly, 1cm posteriorly and 0.8cm inferiorly/laterally. In the oART workflow, artificial intelligence (AI) was utilized during daily plan generation for: organs at risk (OAR) contour delineation; structure-guided target contour propagation; and automated RT plan generation. RTTs were trained to perform the tasks of the oART workflow and a local competency program for RTTs was applied. For the first five patients a radiation oncologist (RO) was present at each treatment, for the second five patients, the RTTs performed the process independently with a RO present if needed. A RO reviewed the images weekly. A comparison of dose delivered using the daily adaptive plan (ADP) versus the original scheduled plan (SP) was performed using Wilcoxon signed rank test, for both PTV coverage and OAR doses. Timing measurements were recorded for the workflow. <h3>Results</h3> Seven male and three female patients with a median age of 79 years (range 66-82) were treated with oART for MIBC. Four patients received concomitant chemotherapy. A total of 186 oART treatments were planned. Of these, the SP was delivered on 10 occasions as a result of technical issues with the online ADP quality assurance. Using the ADP, PTV D98% was increased by a mean of 3.3% (p<0.05), and bowel and rectal V52 Gy were reduced by a mean of 9.1cc (p<0.05) and 2.6% (p<0.05), respectively, compared to the SP. For both the RO and RTT-led workflows, the median adaptive process duration from CBCT acquisition to beam-on was 19 minutes (range 9-35 minutes). <h3>Conclusion</h3> The use of AI has enabled routine clinical implementation of daily CBCT-guided oART which is RTT led. The use of oART increased PTV coverage and reduced OAR doses for MIBC. The development of appropriately trained RTTs to run this service increases the likelihood that this technological advancement can be maintained. Steps to reduce PTV margins are to be explored which should further enhance the dosimetric benefits of oART.