Abstract / Summary
Reproducible delineation of gross tumour volumes is essential for imaging-based target definition in breast cancer, particularly in potential preoperative radiotherapy workflows. However, manual MRI-based segmentation is prone to inter-observer variability. This study explores whether 18 F-FDG PET/CT, through a guideline-based semi-automated segmentation workflow, improves reproducibility for both primary tumour and nodal disease. We performed a retrospective in silico study of 30 patients with stage II–III breast cancer who underwent both contrast-enhanced MRI and 18 F-FDG PET/CT before neoadjuvant chemotherapy. Two independent multidisciplinary expert groups delineated primary tumour (GTV-T) and nodal metastases (GTV-N) using manual MRI and semi-automated PET/CT (fixed 41% SUVmax threshold; EANM-compliant). Volumes were compared using paired t-tests and Bland–Altman analysis. Inter-observer reproducibility between groups was quantified using Dice similarity coefficient (DSC). PET/CT-derived GTV-T volumes were consistently smaller than MRI-based volumes, reaching statistical significance in one observer group ( p = 0.04). For GTV-N, mean volumes were comparable across modalities. However, PET/CT achieved superior inter-observer agreement: mean DSCs were 0.81 (GTV-T) and 0.83 (GTV-N), compared with 0.79 and 0.73 for MRI. Optimal reproducibility (DSC ≥ 0.80) was achieved in 70% of tumours and 80% of nodal volumes with PET/CT, versus 53% and 33% with MRI ( p = 0.01 for GTV-N). PET/CT identified all 44 metastatic lymph nodes with 100% concordance, outperforming MRI (80%). An EANM guideline-based semi-automated ^18F-FDG PET/CT segmentation workflow provides more reproducible GTV delineation than manual MRI, particularly for nodal disease. Its functional target definition and supine acquisition support integration into radiotherapy workflows, including preoperative strategies requiring consistent transfer of target volumes, and may improve standardization in stage II–III breast cancer.