MRI vs CT-based 2D-3D auto-registration accuracy for quantifying shoulder motion using biplane video-radiography

Area of Science:

• Biomedical Engineering
• Radiology
• Orthopedics

Background:

• 3D glenohumeral (GH) joint kinematics are measured using biplane 2D-3D registration.
• Computed tomography (CT) offers high accuracy for 3D bone models, while magnetic resonance imaging (MRI) avoids radiation and visualizes soft tissues.
• The accuracy of MRI-based 2D-3D registration for GH kinematics is not well-established.

Purpose of the Study:

• To develop and evaluate an automatic 2D-3D registration algorithm for both CT- and MRI-based image volumes.
• To compare the humerus and scapula tracking accuracy of CT- vs. MRI-based registration using radiostereometric analysis (RSA) during dynamic biplanar video-radiography.

Main Methods:

• An automatic 2D-3D registration program was developed for CT and MRI data.
• The program was used to quantify humerus and scapula motion during dynamic biplanar video-radiography.
• Accuracy was assessed by comparing CT- and MRI-based registration to radiostereometric analysis (RSA).

Main Results:

• CT-based registration achieved glenohumeral (GH) kinematic accuracy (RMS error) of 0.6-1.0 mm and 0.6-2.2°.
• MRI-based registration yielded GH kinematic accuracy of 1.4-2.2 mm and 1.2-2.6°.
• CT-based registration demonstrated higher accuracy, as expected due to MRI's lower spatial resolution and bone contrast.

Conclusions:

• The developed automatic 2D-3D registration algorithm is effective for both CT and MRI datasets.
• While CT-based registration is more accurate, MRI-based registration provides acceptable accuracy for many clinical research applications.
• This study validates MRI as a viable, radiation-free alternative for quantifying GH joint kinematics.