Development of Prostate Bed Delineation Consensus Guidelines for Magnetic Resonance Image-Guided Radiotherapy and Assessment of Its Effect on Interobserver Variability

Abstract

Purpose: The use of magnetic resonance imaging (MRI) in radiotherapy planning is becoming more widespread, particularly with the emergence of MRI-guided radiotherapy systems. Existing guidelines for defining the prostate bed clinical target volume (CTV) show considerable heterogeneity. This study aimed to establish baseline interobserver variability (IOV) for prostate bed CTV contouring on MRI, develop international consensus guidelines, and evaluate its effect on IOV. Methods and Materials: Participants delineated the CTV on 3 MRI scans, obtained from the Elekta Unity MR-Linac, as per their normal practice. Radiation oncologist contours were visually examined for discrepancies, and interobserver comparisons were evaluated against simultaneous truth and performance level estimation (STAPLE) contours using overlap metrics (Dice similarity coefficient and Cohen's kappa), distance metrics (mean distance to agreement and Hausdorff distance), and volume measurements. A literature review of postradical prostatectomy local recurrence patterns was performed and presented alongside IOV results to the participants. Consensus guidelines were collectively constructed, and IOV assessment was repeated using these guidelines. Results: Sixteen radiation oncologists’ contours were included in the final analysis. Visual evaluation demonstrated significant differences in the superior, inferior, and anterior borders. Baseline IOV assessment indicated moderate agreement for the overlap metrics while volume and distance metrics demonstrated greater variability. Consensus for optimal prostate bed CTV boundaries was established during a virtual meeting. After guideline development, a decrease in IOV was observed. The maximum volume ratio decreased from 4.7 to 3.1 and volume coefficient of variation reduced from 40% to 34%. The mean Dice similarity coefficient rose from 0.72 to 0.75 and the mean distance to agreement decreased from 3.63 to 2.95 mm. Conclusions: Interobserver variability in prostate bed contouring exists among international genitourinary experts, although this is lower than previously reported. Consensus guidelines for MRI-based prostate bed contouring have been developed, and this has resulted in an improvement in contouring concordance. However, IOV persists and strategies such as an education program, development of a contouring atlas, and further refinement of the guidelines may lead to additional improvements.