3D meshless prostate segmentation and registration in image guided radiotherapy.

Area of Science:

• Medical Physics
• Radiology
• Image Processing

Background:

• Image Guided Radiation Therapy (IGRT) is crucial for precise prostate cancer treatment.
• Accurate delineation of organs like the prostate, bladder, and rectum is essential for optimizing IGRT.
• Current methods for organ segmentation can be time-consuming, impacting online treatment delivery.

Purpose of the Study:

• To develop and validate a fast segmentation-registration-segmentation framework for delineating critical organs in Cone Beam CT (CBCT) images.
• To improve the accuracy and efficiency of organ segmentation for online IGRT in prostate cancer patients.
• To enable precise radiation dose coverage while minimizing exposure to adjacent normal tissues.

Main Methods:

• A framework utilizing deformable models, level set formulation, and meshless point sets for automatic segmentation of prostate, bladder, and rectum.
• 3D non-rigid registration of segmented models from planning CT to treatment CBCT images, with global shape prior constraints.
• Refinement of registered models in the CBCT domain using object interactions and image features.

Main Results:

• The developed framework achieved automatic segmentation of the prostate, bladder, and rectum.
• Validation against manual segmentations by a radiation oncologist showed satisfactory convergence.
• The automated segmentation met the speed requirements for online IGRT applications.

Conclusions:

• The proposed fast segmentation-registration-segmentation framework accurately and efficiently delineates critical organs in CBCT images.
• This method enhances the precision of IGRT for prostate cancer, optimizing radiation delivery and organ sparing.
• The framework's speed and accuracy support its integration into online IGRT workflows.