KEY POINTS
- This retrospective in-silico study included 30 women with stage II–III breast cancer who underwent both contrast-enhanced MRI and FDG PET/CT before neoadjuvant chemotherapy.
- Two blinded multidisciplinary expert groups independently contoured the primary tumor and involved nodes. MRI contours were manual, while PET/CT used a standardized semi-automated 41% SUVmax threshold followed by CT-based anatomical verification.
- For primary tumors, inter-observer Dice scores were similar but numerically higher with PET/CT: 0.81 versus 0.79 with MRI. Optimal overlap, defined as Dice ≥0.80, occurred in 70% versus 53% of tumors.
- The larger advantage was in nodal disease. Mean nodal Dice increased from 0.73 with MRI to 0.83 with PET/CT, and Dice ≥0.80 was achieved in 80% versus only 33% of nodal volumes.
- PET/CT identified all 44 metastatic lymph nodes with 100% concordance between observer groups. Complete MRI agreement for nodal disease was achieved in only 80% of patients.
- PET/CT-derived primary tumor volumes tended to be smaller than MRI volumes. In one observer group, mean GTV fell from 14.2 cc with MRI to 10.3 cc with PET/CT (p=0.04).
- Practical workflow may also favor PET/CT because it was acquired supine, closer to the treatment position, whereas diagnostic breast MRI was performed prone. This could reduce additional registration uncertainty in preoperative RT or regional nodal planning.
- The study did not generate RT plans, compare DVHs, validate contours against pathology, or examine local control. The 30-patient single-centre cohort therefore establishes reproducibility rather than target accuracy or clinical benefit.
CLINICAL TAKEAWAY
Standardized FDG PET/CT segmentation may be particularly useful when gross nodal disease must be transferred reliably into a breast RT plan. The improvement in reproducibility is compelling technically, but more reproducible does not necessarily mean more anatomically correct, and prospective dosimetric and pathological validation is still required.