Development and evaluation of an articulated registration algorithm for human skeleton registration

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Radiology

Background:

• Accurate registration of multiple scans is crucial for monitoring bone disease treatment response, particularly for metastatic bone lesions.
• Existing registration methods struggle to account for the natural flexibility and postural variations of the whole-body skeleton.

Purpose of the Study:

• To develop and evaluate an articulated registration algorithm for whole-body human skeleton registration.
• To compare the performance of articulated registration against rigid and deformable methods for skeletal analysis.

Main Methods:

• Whole-body skeletons were extracted from CT images of 16 prostate cancer patients.
• Articulated, rigid, and deformable registration algorithms were applied to pre- and mid-treatment scans.
• Registration performance was assessed using Dice Similarity Index (DSI) and Hausdorff distance (dHausdorff).

Main Results:

• Articulated registration achieved a DSI of 75 ± 2% and dHausdorff of 0.37 ± 0.08 cm, outperforming rigid registration (DSI=59 ± 9%, dHausdorff=0.69 ± 0.20 cm).
• Deformable registration showed similar performance (DSI=77 ± 3%, dHausdorff=0.34 ± 0.08 cm) but deformed over 60% of skeletons, compromising local bone rigidity.
• Articulated registration effectively balanced global flexibility and local bone rigidity.

Conclusions:

• Articulated registration is a superior method for whole-body skeleton registration compared to rigid and deformable techniques.
• This algorithm enables accurate assessment of treatment response for various bone diseases by preserving skeletal integrity.
• The developed articulated registration method holds significant potential for clinical applications in oncology and other bone-related conditions.