3D CT to 2D low dose single-plane fluoroscopy registration algorithm for in-vivo knee motion analysis

Area of Science:

• Medical imaging
• Biomedical engineering
• Radiology

Background:

• Low-dose X-ray fluoroscopy images suffer from noise and blurring, limiting accurate automatic tracking of knee motion.
• Quantum noise is the predominant noise source in low-dose X-ray imaging.
• Preserving anatomical structures like bone is crucial during noise reduction for effective analysis.

Purpose of the Study:

• To develop an accurate multi-modal image registration algorithm for 3D CT to 2D fluoroscopy images of the knee.
• To introduce a novel registration framework incorporating a filtering method to reduce noise and blurring in fluoroscopy images.
• To enhance the accuracy and repeatability of in vivo knee joint motion analysis.

Main Methods:

• A multi-modal image registration algorithm was developed to align 3D CT scans with 2D low-dose fluoroscopy images.
• A new registration framework was employed, featuring a pre-filtering step to mitigate noise and blurring.
• The algorithm was tested on healthy knee fluoroscopy images.

Main Results:

• The proposed algorithm successfully registered 3D CT to 2D noisy and blurred fluoroscopy images.
• The inclusion of a pre-filtering step significantly reduced noise and blurring effects.
• Experimental results demonstrated higher accuracy and repeatability in knee joint motion analysis.

Conclusions:

• The developed image registration algorithm effectively addresses noise and blurring in low-dose fluoroscopy.
• The pre-filtering approach within the registration framework improves the reliability of knee motion analysis.
• This technique offers a promising solution for more accurate in vivo knee joint motion tracking.