Deformable image registration for adaptive radiotherapy with guaranteed local rigidity constraints

Area of Science:

• Medical Physics
• Radiotherapy Technology
• Image Analysis

Background:

• Deformable image registration (DIR) is crucial for radiotherapy, but existing methods often fail to account for varying tissue properties, leading to inaccurate deformations.
• Improving DIR plausibility in male pelvic adaptive radiotherapy requires models that consider local anatomical rigidity.

Purpose of the Study:

• To integrate a local rigidity deformation model into a DIR algorithm for enhanced accuracy in male pelvic radiotherapy.
• To improve the physical plausibility of DIR by enforcing rigid behavior in specific anatomical structures.

Main Methods:

• A DIR framework was extended with constraints for locally rigid deformation of delineated structures (bones, prostate).
• The algorithm allowed surrounding tissues to deform elastically while restricting specified regions to rigid transformations.
• The approach was tested on ten male pelvic CT/CBCT datasets and compared against Varian SmartAdapt deformable registration (VSA).

Main Results:

• The new approach achieved comparable accuracy (Dice Similarity Coefficient: 0.87) to VSA (0.86) but with significantly fewer deformation grid foldings (e.g., 0.08% in bladder for REG2 vs. 1.74% for VSA).
• Rigidity constraints successfully maintained bone and prostate integrity, unlike REG0 and VSA which showed elastic bone deformations.
• The method achieved clinically feasible runtimes (26.2s for REG1, 31.1s for REG2) comparable to non-rigid approaches.

Conclusions:

• The constrained DIR approach provides physically plausible deformations in the male pelvic region with guaranteed rigidity of selected structures.
• The method achieves accuracy comparable to VSA while offering improved robustness against grid foldings.
• The algorithm demonstrates potential for enhancing adaptive radiotherapy through accurate and plausible image registration.