ESTRO 2025 - Abstract Book

S2560

Physics - Autosegmentation

ESTRO 2025

Figure 2, derived from metrics calculated using data from the TPS database, highlights site-specific trends such as: • Breast: Large number of complete modifications. • Urology: Frequent but minor corrections. • Lung: Infrequent and minor adjustments. • Head and Neck: The highest variability due to the numerous OARs involved.

In terms of organizational efficiency, the software achieved time savings of 60%–90% for generating complete OAR structure sets, depending on the anatomical site. Additionally, workflow dead times were eliminated, leading to anticipated improvements in overall workflow efficiency. Conclusion: This study represents a comprehensive correlation of qualitative Likert survey data with large-scale contour modification analysis over a year of clinical use. While auto-segmentation demonstrated the capability to produce high-quality contours, human oversight and adjustments remained essential to ensure clinical safety and maintain high standards of care. The implementation of AI-RAD Companion Organs RT resulted in substantial workflow improvements, with clinician satisfaction aligning with expectations reported in prior studies [2-3].