ESTRO 2024 - Abstract Book

S3161

Physics - Autosegmentation

ESTRO 2024

the one used by Pera et al [ 4 ]. To avoid bias, this survey was conducted without knowing the origin of each set of structures.

Table 1

Likert scale

Description

Clinally aceptable for radiotherapy planning

3

Agree

2

Neither Agree

May be used but edits are neccesary

1

Disagree

Unusable and redone manually

Results:

The distribution of the VS parameter per patient is shown in Figure 1 for each OAR and system. Higher means were observed for OAR generated by Siemens, around 0.98 vs 0.95 for Raystation. This was predictable because the clinical volumes were based on the first system. The VS for the rectum was lower for both systems, with an average of 0.93. Additionally, outliers were observed in the bladder and femoral heads, which were identified in patients with very small bladders or femoral prostheses. Notably, Siemens performed better in cases involving femoral prostheses, as indicated by the VS values. Survey results contradicted some of the above data. Figure 2 reveals that, according to radiation oncologists' opinions, Raystation excelled in bladder and rectum contouring, while Siemens performed better for femoral heads. Most agreed that Raystation was more effective in identifying bladder tumor infiltration but the femoral heads are contoured a bit short. Both systems were considered to contour the rectum excessively large, without distinguishing between the rectum and sigma. As an extra, the auto-contours of the 8 cases best rated by the survey have been chosen and the impact they would have on the dosimetry, [ 5 ] planned by dosimetrists with more than 10 years of experience and used in clinical treatment, has been evaluated. For V37 of the rectum and bladder, differences of up to 6% and 3%, respectively, have been observed for both systems. On average, dosimetric changes have been found to be smaller in Raystation contours. In the case of femoral heads there are no appreciable differences in V30 for neither.