ESTRO 2024 - Abstract Book

S169

Brachytherapy - GI, paediatric, miscellaneous

ESTRO 2024

101

Poster Discussion

Integration of Artificial Intelligence in Interstitial HDR Brachytherapy for Liver Tumors

Hathal Haddad 1 , Horst Hermani 2 , Michael Pinkawa 1

1 Robert Janker Clinic, Radiation Oncology, Bonn, Germany. 2 Robert Janker Clinic, Medical Physics, Bonn, Germany

Purpose/Objective:

Artificial intelligence (AI) integration in radiation therapy has seen a notable upswing in recent years. This study explores AI's application in contouring during interstitial high-dose-rate (HDR) brachytherapy for primary and secondary liver tumours

Material/Methods:

The study encompassed 25 patients, comprising 37 hepatic lesions of primary and secondary liver tumours. Interstitial HDR brachytherapy was employed for treatment, with the Brachytherapy Applicators strategically placed under CT guidance. A Planning CT with a 1.25 mm slice thickness and contrast medium was subsequently conducted. Organs at risk (OAR) were delineated by both an experienced Radiation Oncologist and AI-powered autocontouring software (Limbus Contour v.1.5.0) using a PC equipped with an Intel® Xeon® E5-2620V3 processor (CPU) and an NVIDIA Quadro K620 graphics card (GPU) with 2GB GDDR3. The contoured organs at risk included the liver, kidneys, stomach, intestine, heart, oesophagus, and lungs. The time taken for contouring the organs at risk was compared.

Results:

The study revealed various types of tumours, as depicted in Figure 1. The experienced radiation oncologist took an average of 17.08±0.61 minutes for contouring. Utilizing CPU calculations with the software, the mean time was reduced to 3.71±0.45 minutes. When employing the GPU, the calculation time further decreased to 2.24±0.23 minutes. The disparity in contouring times was statistically significant (p < 0.001). Stomach contouring was necessary for 21 metastases, while right kidney and intestine contouring was required for 26 and 33 metastases, respectively.