ESTRO 2024 - Abstract Book

S5995

RTT - Treatment planning, OAR and target definitions

ESTRO 2024

Lucrezia Bernabucci 1 , Antonella Martino 1 , Mariangela Massaccesi 1 , Francesco Beghella Bartoli 1 , Rosa Autorino 1 , Stefania Teodoli 2 , Enrico Rosa 3 , Silvia Chiesa 1 , Vincenzo Frascino 1 , Stefania Manfrida 1 , Maria Antonietta Gambacorta 1 , Luca Indovina 2 , Nicola Dinapoli 1 1 Fondazione Policlinico Universitario A. Gemelli IRCCS, Radiotherapy, Rome, Italy. 2 Fondazione Policlinico Universitario A. Gemelli IRCCS, Medical Physics, Rome, Italy. 3 Università Cattolica del Sacro Cuore, Medical Physics, Rome, Italy

Purpose/Objective:

Accurate delineation of organs at risk (OARs) is essential for effective treatment planning in brain radiotherapy. This study aims to assess the role of MIM, an autocontouring software, in the segmentation of OARs in the brain. We compared three datasets: contours generated exclusively by MIM, contours generated by MIM and subsequently manually adjusted, and contours manually drawn by experts.

Material/Methods:

The study involved a total of 23 patients undergoing brain radiotherapy. OARs, including optic nerves, chiasm, lens, brainstem, and cochlea, were delineated using three methods: MIM's autocontouring, MIM's autocontouring with manual refinement, and manual contouring by experts. We evaluated the accuracy of OAR volumes using the Dice Similarity Coefficient (DSC). The DSC values were compared among the three contouring methods using the paired t test. A statistical significance level was set at p < 0.05. We also examined time-saving implications.

Results:

The initial contours generated by MIM demonstrated acceptable segmentation of brain OARs. However, subsequent manual adjustments significantly enhanced the accuracy and consistency of the contours. These manual refinements accounted for anatomical variations and subtle boundaries that the automated software might overlook. Quantitative analysis revealed a reduction in volume disparities when comparing the manually adjusted MIM contours with those drawn by experts (see the table). The time-saving benefit was evident in the contouring process using MIM software (average time: 6 minutes, range: 2-10) compared to manual contouring (average time: 28 minutes, range: 20-40).

Conclusion:

MIM's autocontouring software serves as a valuable starting point for the segmentation of brain OARs in radiotherapy treatment planning. However, to achieve optimal accuracy, manual adjustments by experienced clinicians are indispensable. We recommend a combined approach, incorporating automated software and expert knowledge, to attain reliable and accurate OAR segmentation in brain radiotherapy. This study underscores the significance of integrating automated tools with manual intervention to ensure high-quality treatment planning.

Keywords: autocontouring software, brain cancer