Multiple organ segmentation framework for brain metastasis radiotherapy.

Area of Science:

• Medical imaging
• Radiotherapy
• Computational anatomy

Background:

• Radiotherapy is a primary treatment for brain metastases, necessitating precise delineation of organs-at-risk (OARs) to prevent radiation-induced toxicity.
• Challenges in OAR delineation include organ size imbalance, unclear boundaries, and complex anatomy, hindering accurate treatment planning.

Purpose of the Study:

• To develop a novel cascaded multi-OAR segmentation framework, OAR-SegNet, to address the challenges in OAR delineation for radiotherapy.
• To improve the accuracy and efficiency of segmenting organs-at-risk in medical imaging for brain metastases treatment.

Main Methods:

• Introduced OAR-SegNet, a two-level segmentation framework consisting of an Anatomical-Prior-Guided network (APG-Net) and a Point-Cloud-Guided network (PCG-Net).
• APG-Net utilizes multi-view segmentation and a deep prior loss guided by anatomical knowledge for initial segmentation of all organs.
• PCG-Net refines segmentation of smaller organs using mini-segmentation and point-cloud alignment, incorporating deep prior features.

Main Results:

• The proposed OAR-SegNet framework demonstrated superior performance in OAR segmentation compared to existing state-of-the-art methods.
• The cascaded approach effectively handled challenges like imbalanced organ sizes and ambiguous boundaries.
• Experimental results validated the framework's capability in precise OAR delineation.

Conclusions:

• OAR-SegNet offers a significant advancement in automated OAR delineation for radiotherapy treatment planning.
• The framework's ability to accurately segment organs-at-risk can lead to reduced radiation toxicity and improved patient outcomes.
• This novel approach provides a robust solution for complex segmentation tasks in medical imaging.