ESTRO 2024 - Abstract Book

S3792

Physics - Image acquisition and processing

ESTRO 2024

Four observers, two experienced radiotherapy technologists and two experienced radiation oncologists, evaluated the pseudo-anonymized CBCT images in a random, blinded fashion with the planning CT image including predefined structures shown next to it as a reference. For each CBCT image, visibility of the prostate, seminal vesicles, bladder and rectum was scored on a scale from 1-5, with 1 defined as not visible and 5 defined as visibility equal to planning CT. In addition, observers annotated each axial slice where delineation of these structures was difficult. Lastly, observers indicated (yes/no) whether the CBCT image quality was sufficient for online adaptive radiotherapy. Overall, a CBCT image was considered of sufficient quality if at least three out of four observers scored ‘yes’. For each CBCT image pair, the differences between the structure scores and the difference in the number of annotated axial slices with poor visibility was calculated per observer for individual structures. Per structure and per patient, difference scores were averaged over the three CBCT imaging pairs and four observers. A Wilcoxon signed rank test with Bonferroni correction for multiple testing was used to test for deviations from zero, and thereby differences between HyperSight CBCT imaging and conventional CBCT imaging.

Results:

Differences between structure scores, averaged per patient, are shown in figure 2. HyperSight CBCT imaging performed significantly better than conventional CBCT imaging for all four structures (p-values: <0.001). The structure scores increased from a median of 3 for conventional CBCT to 5 for HyperSight CBCT for prostate, from 4 for conventional CBCT to 5 for HyperSight CBCT for bladder, from 3 for conventional CBCT to 5 for HyperSight CBCT for rectum and from 3 for conventional CBCT to 5 for HyperSight CBCT for seminal vesicles.