Multidisciplinary approach to target volume delineation in locally recurrent rectal cancer: An explorative study

Abstract

Background and purpose

Interobserver variation (IOV) in locally recurrent rectal cancer (LRRC) delineations is large, possibly because of different interpretations of imaging. An explorative study was performed to investigate the benefit of additional delineations by expert radiologists.

Materials and methods

14 cases of LRRC were delineated on planning CT by 8 radiologists (RADs) to construct a median and total radiology contour, followed by 12 radiation oncologists (ROs), without (GTV−) or with (GTV+) the additional contours. IOV was calculated separately for RADs, GTV− and GTV+. The following metrics were used: the Surface Dice Similarity Coefficient (SDSC), Dice similarity coefficient (DSC), and Hausdorff Distance at the 98th percentile (HD98%). The median SDSC, DSC, and HD98% of GTV− and GTV+ were compared. Sub-analyses of IOV in different recurrence types were performed.

Results

Median SDSC significantly improved from GTV− to GTV+ overall, but a significant benefit could not be proven in individual cases. Additional radiological input consistently improved all parameters in 4/14 cases (29 %). Geographical miss occurred after radiological input in 7 %. Subgroup analyses show large IOV in mainly fibrotic and intraluminal recurrences. Little IOV is seen in solitary nodal recurrences.

Conclusion

This study highlights target volume delineation challenges in LRRC. Overall, radiological input reduced IOV amongst ROs in target volume delineation for LRRC. Large differences do however exist amongst recurrence types. A standard terminology for LRRC and close collaboration between radiologists and radiation oncologists seems necessary to reduce IOV and improve quality of care.