Automatic target volume segmentation for offline adaptive head-and-neck radiotherapy.

Area of Science:

• Medical Physics
• Radiotherapy Oncology
• Medical Imaging

Background:

• Offline adaptive radiotherapy requires accurate target volume delineation.
• Manual contouring is time-consuming and prone to inter-observer variability.
• Deformable image registration (DIR) offers automated solutions for contour propagation.

Purpose of the Study:

• To evaluate the clinical utility of Smartfuse, a DIR algorithm, for automatic target volume propagation in head-and-neck radiotherapy.
• To assess the geometric and dosimetric accuracy of Smartfuse-propagated contours compared to manual delineations.

Main Methods:

• Ten head-and-neck cancer patients undergoing offline adaptive radiotherapy were included.
• Manual delineation of gross tumor volume (GTV) and clinical target volume (CTV) was performed on initial (CT_i) and re-planning (CT_R) CT scans.
• Smartfuse propagated contours from CT_i to CT_R; geometric agreement was assessed using Dice Similarity Coefficient (DSC) and Hausdorff distance (HD₉₅).
• Dosimetric evaluation compared plans based on DIR-propagated targets (PTV_DIR) versus manually delineated targets (PTV_RO).

Main Results:

• Median geometric agreement showed a Dice Similarity Coefficient (DSC) of 0.86 and a 95th percentile Hausdorff distance (HD₉₅) of 4.0 mm.
• Dosimetric analysis revealed minimal relative differences in D95% for GTV+CTV (-0.6%) and PTV (-2.1%).
• While all GTV+CTV met V95% ≥ 95%, only 61% of PTV_RO achieved this with DIR plans, indicating potential coverage discrepancies.

Conclusions:

• Smartfuse demonstrates potential for efficient automatic propagation of target volumes in head-and-neck radiotherapy.
• Medical review of auto-propagated volumes is essential.
• Reliance solely on automatically generated planning target volumes may lead to dosimetric inaccuracies.