ESTRO 2024 - Abstract Book

S228

Brachytherapy - Gynaecology

ESTRO 2024

Interstitial brachytherapy is a primary treatment modality for cervical cancer, involving the insertion of radioactive sources, such as needles, into or near tumor tissues. The proper choice of needle number and precise placement of these needles are critical to optimize the dose distribution, the treatment effectiveness, and the protection of healthy tissues. However, due to patient-specific variations in anatomy, a high level of training experience and expertise are required to navigate needles. Therefore, this study aims to develop a virtual reality (VR) training system for interstitial brachytherapy, enabling practitioners to place needles within virtual tissues and automatically evaluate their alignment with the real-world reference needle positions.

Material/Methods:

We retrospectively obtained CT scans from two cervical cancer patients who underwent interstitial brachytherapy in our institution. Radiotherapy structures, including gross tumor volume (GTV), organs at risk (OARs), and the patient's body, were delineated with Ocentra 4.3 software, and the positions of the real needles were exported as baseline for reference. By intentionally generating a sphere at the model's origin, we established a reference point to bridge the virtual and the real-world coordinates, enabling accurate simulations of needle placement and subsequent reconstruction. The CTs and radiotherapy structures were then pre-processed in 3D Slicer 5.2.2. Additionally, a virtual reality (VR) system was developed by converting the model to FBX format in Maya 2022 and building the background scenario in Unreal Engine 5.0. The training was conducted in the Pico 4.0 system, enabling physicians to simulate needle placement and record their positions within the VR tissues, which were later reconstructed to their coordinates in real world by Matlab R2014A. Finally, to evaluate the needle distribution, we mapped the positions of both baseline and simulated needles to binary masks and calculated the Hausdorff distance at the 95th percentile (HD95). The scenograph of the VR training system was displayed in Fig. 1.

Results:

The binary mask distributions of the baseline and the simulated needles from a randomly selected CT slice of the two patients were displayed in Fig.2, using small circles with a radius of 1.0 mm to indicate the needle positions. The two testing cases had maximum tumor diameters of 19.5 mm and 40.1 mm at the axial plane, with 2 and 5 interstitial needles, respectively. The resulting HD95 values were 1.9 mm and 2.8 mm, indicating the VR training system could effectively simulate the needle placement within tissues and match with the real-world positions. Nevertheless, the existing discrepancies may be attributed to real-time adjustments, organ movement, and image-guided feedback during the actual needle placing process.