Quantitative assessment of inter-observer consistency in target volume delineation following breast-conserving surgery and mastectomy.

BACKGROUND

Consistent clinical target volume (CTV) delineation is vital for breast cancer radiotherapy. While organs-at-risk definitions are increasingly automated, manual CTV delineation of complex post-surgical volumes remains highly observer-dependent, potentially compromising treatment homogeneity.

OBJECTIVE

To quantify inter-observer discrepancies in postoperative breast cancer CTV contouring among regional oncologists and to qualitatively analyze the reasons for these contouring discrepancies to promote standardization and quality assurance.

METHODS

Based on CT datasets of two representative post-surgical breast cancer cases (BCS and MRM), 22 frontline radiation oncologists from regional municipal hospitals independently contoured the required CTVs. Inter-observer contouring discrepancies across these manual target volumes were quantitatively evaluated via Dice similarity coefficients (DSC) and Hausdorff distances (HD) against a consensus reference standard established through a two-expert consensus workflow.

RESULTS

A total of 20 completed datasets, each for both the BCS and MRM cases, were successfully retrieved and analyzed. The large volumes achieved the highest overlap (CTVp_breast and CTVp_thoracic wall DSC mean: 0.80 ± 0.07 and 0.63 ± 0.08, respectively), though mean 95% Hausdorff distances (HD95) were elevated at 42.84 ± 50.62 mm and 45.73 ± 60.44 mm. Post-BCS focal volumes (CTV_tb and CTV_boost) showed intermediate overlap (both DSC mean: 0.56) but restricted boundary deviations (mean HD95: 11.51 mm and 19.94 mm). For nodal stations, consistency was moderate for the supraclavicular volume (CTVn_L4 DSC mean: 0.57 ± 0.15, HD95: 26.21 ± 15.08 mm) but dropped substantially for axillary Level III (CTVn_L3 DSC mean: 0.38 ± 0.26, HD95: 37.95 ± 37.94 mm) and internal mammary nodes (CTVn_IMN DSC mean: 0.42 ± 0.14, HD95: 42.67 ± 25.70 mm).

CONCLUSION

Delineation consistency is acceptable for large volumes (CTVp_breast and CTVp_thoracic wall) but poor for regional nodal stations, demonstrating that standardized anatomical training is crucial to improve radiotherapy uniformity.