Multi-needle Localization for Pelvic Seed Implant Brachytherapy based on Tip-handle Detection and Matching

Area of Science:

• Medical Imaging
• Computational Anatomy
• Radiotherapy Physics

Background:

• Intraoperative CT imaging is vital for precise seed placement in pelvic brachytherapy.
• Challenges include poor image contrast and needle adhesion, hindering accurate multi-needle localization.
• Robust localization is essential for optimizing radiation dose delivery and patient outcomes.

Purpose of the Study:

• To develop and validate a novel approach for accurate multi-needle localization in intraoperative CT images.
• To address the limitations of current methods in detecting and reconstructing needle paths.
• To improve the precision of seed placement in pelvic brachytherapy.

Main Methods:

• A novel approach reframing needle localization as a tip-handle detection and matching problem.
• An anchor-free network based on HRNet for multi-scale feature extraction and detection of needle tips and handles.
• A greedy matching and merging (GMM) method to associate detected tips and handles and reconstruct 3D needle paths.

Main Results:

• The proposed method achieved superior performance in needle localization compared to segmentation-based baselines.
• Higher precision and F1 scores were demonstrated on a dataset of 100 patients.
• The method effectively reconstructs 3D needle paths, overcoming challenges of poor contrast and needle adhesion.

Conclusions:

• The novel tip-handle detection and matching approach offers a robust and accurate solution for multi-needle localization in intraoperative CT.
• This advancement has the potential to significantly enhance seed placement accuracy in pelvic brachytherapy.
• The developed method provides a valuable tool for complex clinical scenarios in radiotherapy.