Main Session
Sep
29
PQA 05 - Physics
3104 - Characterizing Contouring Uncertainty across Clinical Specialties in CBCT-Guided Online Adaptive Partial Breast Irradiation
Presenter(s)
Udbhav Ram, BS - University of Alabama at Birmingham, Birmingham, AL
U. S. Ram1, J. Duan2, R. A. Cardan1, A. Robbins3, C. Dobelbower1, C. Speers4, C. Stanley1, D. N. Stanley1, H. Boggs5, J. Lacy3, K. S. Keene5, M. Wine3, M. Soike4, N. Viscariello1, N. Lima3, M. Tyler2, C. Cardenas1, and J. A. Pogue1; 1University of Alabama at Birmingham, Birmingham, AL, 2Department of Radiation Oncology, University of Alabama at Birmingham, Birmingham, AL, 3The University of Alabama at Birmingham, Birmingham, AL, 4University of Alabama at Birmingham, Department of Radiation Oncology,, Birmingham, AL, 5University of Alabama at Birmingham Department of Radiation Oncology, Birmingham, AL
Purpose/Objective(s):
Online adaptive APBI is gradually being adopted at academic centers using varied staffing models, where GTV definition on CBCT may be performed by different members of the multidisciplinary team. The goal of this study was to quantify profession-dependent contouring variability in CBCT-based GTV editing and its downstream impact on PTV margin requirements.Materials/Methods:
Fourteen early-stage breast cancer patients treated with CBCT-guided online adaptive APBI (30 Gy/5) were included in this retrospective analysis. Fourteen observers (5 radiation oncologists, 3 physicists, 3 dosimetrists, 3 therapists) independently edited deformable image registration (DIR)-propagated lumpectomy cavity GTVs on fractions 1 and 5. A consensus target was generated via 3/5 majority voting (MV) on physician GTVs. CTVs were derived using a 1 cm expansion with anatomic cropping, and PTVs were generated using 1–5 mm isotropic expansions. Agreement was quantified using consensus CTV coverage (true positive rate) and volume ratio (relative expansion).Results:
0 mm PTV overlap showed large outliers across professions (some cases <70% vs MV contours). Coverage increased steeply with small margins; by 3 mm, median physician-consensus coverage was ~100% across groups. With 3 mm expansion, physician-edited workflows had the lowest median volume increase (~37%) while unedited autocontours needed more (~51%), implying more normal tissue treated for similar coverage. To reach =95% MV coverage in 90% of cases, physicians required 3 mm, whereas others needed larger margins (~5 mm).Conclusion:
In this single-institution study, profession-dependent IOV in CBCT-based GTV editing impacted PTV margin requirements; standard margins (~3 mm) restored consensus volume ovelrap, whereas unedited autocontours and some non-physician edits required larger margins (up to ~5 mm), increasing treated volume for similar consensus overlap. A risk-stratified approach is supported, using profession-aware margins and competency-based credentialing, supporting physician contouring for cases where minimizing treated volume is critical. Table 1. Profession-dependent margin, volume expansion, and contouring time (n=14 patients; 14 observers)| Observer Group (n) | Margin to achieve =95% CTV coverage in 90% of cases (mm) | Volume expansion at 3-mm PTV margin (%) | Median contouring time (min) |
| Radiation oncologists (5) | 3 | 37.1 | 1.92 |
| Medical physicists (3) | 5 | 39.2 | 3.25 |
| Dosimetrists (3) | 3 | 45.5 | 6.36 |
| Radiation therapists (3) | 1 | 45.7 | 5.48 |
| Unedited autocontours (DIR) | 2 | 51.0 | — |