Clinically Oriented Contour Evaluation Using Dosimetric Indices Generated From Automated Knowledge-Based Planning.

Area of Science:

• Radiation Oncology
• Medical Physics
• Radiotherapy Planning

Background:

• Geometric indicators for contouring accuracy in radiation therapy often lack clinical context.
• Manual contouring variations can confound treatment plan evaluation.
• Objective tools are needed to assess contouring quality in radiotherapy.

Purpose of the Study:

• To assess the feasibility of using automated knowledge-based planning for objective contour evaluation.
• To determine the correlation between geometric indices and dosimetric endpoints.
• To report the dosimetric impact of variations in head and neck target and organ-at-risk (OAR) volume contours created by resident physicians.

Main Methods:

• Twenty-two resident physicians contoured head and neck structures for a nasopharyngeal cancer case.
• Automated knowledge-based planning generated 67 treatment plans using resident and gold-standard contours.
• Geometric indices (Dice, Hausdorff, volume difference, centroid distance) and dosimetric indices were calculated and correlated (R²).

Main Results:

• A weak and inconsistent correlation (R² < 0.2 for 61%) was found between geometric and dosimetric indices.
• No single geometric index consistently predicted dosimetric impact, preventing the establishment of clinical acceptability thresholds.
• Geometric indices showed high rates of false positives/negatives in classifying dosimetric impact, and significant inter-resident contour variation was observed.

Conclusions:

• Contour variations among resident physicians significantly impact dosimetric endpoints in head and neck radiotherapy.
• Resident education in anatomy delineation is critical for improving contouring accuracy.
• Dosimetric indices from automated planning should supplement geometric indices for contour evaluation in radiotherapy studies.