ESTRO 2025 - Abstract Book

S2202

Interdisciplinary – Education in radiation oncology

ESTRO 2025

Purpose/Objective: Expertise in the contouring of anatomical organs is required to verify and correct the auto-contouring results generated by artificial intelligence (1). This study aims to evaluate the effectiveness of a guideline-based training platform used to improve contouring skills.

Material/Methods: MVision Guide was used in this research. In four anatomical regions (brain, breast, pelvis male and female, and head&neck), eleven, seven, six, four and six organs, respectively were contoured by RTTs and young residents. Contours of experts on five CT series were selected as references, and tests and practices were performed according to the timeline shown in Figure 1.

The first three CT series were used to test and practice, and the fourth one only to test. Contouring on the fifth CT series was performed only at the beginning (Test_1) and at the end (Test_3). To quantify structure set similarities, volume difference, Dice similarity coefficient (DSC), surface DSC (s-DSC), and Hausdorf distance (HD95) were used. Results: Figure 2 shows an example of an improvement in the contouring of six organs in the male pelvis region by a young resident.

In the brain region, an average DSC increase of 32%, 73%, 54%, and 23% was found in the brainstem, left cochlea, right cochlea, and optic chiasm, respectively. No improvement was observed in the eyes and spinal canal, but the original DSC values were high. In the breast region, a slight increase was seen in the LAD and thyroid gland. The average DSC of the bladder in the male pelvis region was the same, 0.97 in Test_1 and Test_4. For prostate and seminal vesicles, an increase in the DSC of 7% and 42%, respectively was obtained between Test_1 and Test_4. In the female pelvis region, the bladder, rectum, and utero-cervix had a DSC value higher than 0.85 in Test_1, so no