ESTRO 2025 - Abstract Book

S2449

Physics - Autosegmentation

ESTRO 2025

As a second example, a plot of clinical contour volume versus deep learning contour volume for femur heads shows that manual adjustments tend to reduce the contour volume.

Conclusion: The dashboard collects data automatically without user intervention. It helps answer questions about DLS performance in clinical practice, such as: • Did a new CT protocol for prostate patients impact autosegmentation quality? • Do clinicians tend to make less manual adjustments over time, indicating automation bias? • Which clinicians make frequent manual adjustments, and which make few? Which organs require more manual adjustments? • Is there a performance bias with respect to tumor volume or patient gender? References: 1 Kiser KJ, Barman A, Stieb S, Fuller CD, Giancardo L. Novel Autosegmentation Spatial Similarity Metrics Capture the Time Required to Correct Segmentations Better Than Traditional Metrics in a Thoracic Cavity Segmentation Workflow. J Digit Imaging. 2021 Jun;34(3):541-553. doi: 10.1007/s10278-021-00460-3. Epub 2021 May 23. PMID: 34027588; PMCID: PMC8329111 2 Vaassen F, Hazelaar C, Vaniqui A, Gooding M, van der Heyden B, Canters R, van Elmpt W. Evaluation of measures for assessing time-saving of automatic organ-at-risk segmentation in radiotherapy. Phys Imaging Radiat Oncol. 2019 Dec 17;13:1-6. doi: 10.1016/j.phro.2019.12.001. PMID: 33458300; PMCID: PMC7807544. Keywords: Quality Assurance, Monitoring, QA Dashboard