Quantification of organ motion based on an adaptive image-based scale invariant feature method.

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Computer Vision

Background:

• Image-guided radiotherapy (IGRT) relies on accurate organ motion quantification for treatment planning.
• Internal organ motion presents challenges in radiotherapy due to its impact on dose delivery.
• Automatic landmark detection is crucial for precise motion analysis in IGRT.

Purpose of the Study:

• To develop an automated approach for localizing anatomical landmarks in IGRT image series.
• To quantify nonrigid motion of anatomical and pathological structures during radiotherapy.
• To utilize local image contrast for improved landmark detection and motion description.

Main Methods:

• An adaptive Scale Invariant Feature Transform (SIFT) was developed by integrating 3D SIFT with local image contrast.
• The method's robustness and invariance to deformable transforms were validated using CT phantom studies.
• The adaptive SIFT was applied to a 4D CT lung dataset, with manual identification serving as ground truth for validation.

Main Results:

• Phantom studies demonstrated high robustness and invariance to deformable transformations with sub-voxel matching errors.
• Automated motion detection of breathing motion in 4D CT datasets was achieved with high accuracy.
• The adaptive SIFT method showed reduced residual errors compared to standard SIFT, comparable to expert manual identification.

Conclusions:

• Adaptive SIFT shows significant potential as an automated tool for landmark detection in deformable image registration (DIR) and internal motion quantification.
• The method facilitates accurate motion description for radiotherapy treatments.
• Future research will focus on optimizing computational efficiency and extending applicability to diverse anatomical sites and imaging modalities.