Quantitative Approaches To Locally Advanced Rectal Cancer Gtv And Subvolume Contouring With Diffusion-Weighted Mri: Implications For Mri-Guided Boost Strategies

To evaluate interobserver agreement between observers for T2-weighted (T2w) and diffusion-weighted MRI (DWI) contours of locally advanced rectal cancer; and to assess the feasibility of manual and semi-automated delineation of restricted diffusion tumor subvolumes for the purposes of radiotherapy dose escalation. Twenty cases of locally advanced rectal cancer were reviewed by two radiation oncologists and two radiologists. Contours of gross tumor volume (GTV) on T2w, DWI and coregistered T2w/DWI (‘CoReg’) were delineated by each observer and a consensus contour for each session was generated using the Simultaneous Truth and Performance Level Estimation (STAPLE) algorithm. Observer-STAPLE comparisons were made using Dice Similarity Coefficient (DSC), mean distance to agreement (MDA) and other metrics of interobserver variability (IOV). Subvolumes of restricted diffusion within CoReg GTVs were manually delineated by observers and created semi-automatically using thresholding of apparent diffusion coefficient centile (ADC) values for comparison. IOV measures for each contouring session were statistically compared with Wilcoxon Signed-Rank tests (p < 0.05). T2w GTV contours demonstrated significantly (p < 0.05) better performance across most IOV metrics than DWI and CoReg contours. Observers were able to delineate subvolumes of restricted diffusion in all cases with moderate agreement. Semi-automated segmentation based on the 40th centile intratumoral ADC value demonstrated best performance across the cohort, but agreement with consensus manual subvolume delineations was moderate to poor. Errors due to inaccurate image registration and luminal motion between T2w and DWI acquisitions contributed to variability in CoReg and subvolume delineations. A validation set of cases with optimized patient preparation and image acquisition processes demonstrated improved agreement and better performance of semi-automated delineations. Contours based on coregistered T2w and DW-MRI imaging could be used for delineation of biologically relevant tumor subvolumes within anatomically accurate GTVs. Semi-automated delineation based on intratumoral ADC thresholds may be a viable way to standardize subvolume delineation if there is accurate registration between imaging acquisitions. This method could be applied to identify relevant treatment subvolumes for radiotherapy dose escalation protocols in locally advanced rectal cancer.Abstract 3348; TableMeasure of agreementT2wDWICoregistered T2w/DWISubvolume (manual)Subvolume (auto, 40th centile ADC)Intraclass Correlation Coefficient0.910.740.560.32-Generalized Conformity Index0.5330.5100.4480.318-Dice Similarity Coefficient0.8180.8040.7680.6660.580Jaccard Index0.7040.6890.6440.5310.418Hausdorff Distance (mm)9.7612.7612.9412.7415.15Mean Distance to Agreement (mm)1.191.662.081.922.44 Open table in a new tab