A novel flexible framework with automatic feature correspondence optimization for nonrigid registration in...

Area of Science:

• Medical Physics and Imaging
• Radiotherapy and Oncology
• Image Registration and Computational Anatomy

Background:

• Radiation therapy advancements have improved cancer treatment but are limited by anatomical changes during treatment.
• Accurate modeling of patient anatomy is crucial for exploiting advanced radiotherapy techniques.
• Nonrigid image registration is a key method for accounting for anatomical variations.

Purpose of the Study:

• To design, implement, and validate a novel nonrigid registration framework for radiotherapy.
• To improve upon the thin plate spline robust point matching (TPS-RPM) algorithm for more accurate and coherent organ registration.
• To enable the simultaneous registration of multiple structures with user-defined constraints.

Main Methods:

• An enhanced TPS-RPM algorithm was developed, incorporating a novel correspondence filter for simultaneous multi-structure registration.
• The framework supports the inclusion of user-defined landmarks, lines, and surfaces.
• Parameters were optimized across head and neck, prostate, and cervix cases, with control points generated from segmented organs.

Main Results:

• The optimized framework achieved transformation errors below 1 mm across all tested sites.
• Inclusion of anatomical features reduced residual distances significantly, e.g., from 1.5 mm to 0.8 mm for head and neck.
• The method produced anatomically coherent transformations without compromising registration accuracy.

Conclusions:

• The developed nonrigid registration framework effectively handles simultaneous registration of multiple organs with varying complexities.
• This tool enhances the precision of radiotherapy by accurately modeling anatomical changes.
• The framework offers a powerful solution for improving radiotherapy planning and delivery.