Automatic lumbar motion analysis based on particle filtering

Abstract

Spinal motion is produced by complex coordination of nerves and muscles and is constrained by vertebral structure. The observation and measurement of lumbar motion is of great value for clinical diagnosis and surgical plan of lumbar disorders. Digitalized Video Fluoroscopy (DVF) is the most suitable one to image the spine motion but it is quite time consuming. This paper proposes an automatic lumbar motion analysis system (ALMAS) with particle filtering technology. The automatically vertebral tracking for motion analysis was utilized with a friendly-interface, which provides a window for users to process the acquired DVF sequence and to analyze the tracking results. A set of simulation vertebra image were used to evaluate the performance and accuracy of this system. In simulated sequence, the maximal difference is 1.3 mm in translation and 1ͦ in rotation angle. The error is small in x- and y- translation (fiducial error: 2.4%, repeatability error: 0.5%) and in rotation angle (fiducial error: 1.0%, repeatability error: 0.7%). The ALMAS can still track the sequence contaminated by noise with the density ≤ 0.5. Besides, the results demonstrate that the data from the auto-tracking algorithm shows a strong correlation with the actual measurement and that the ALMAS is highly repetitive. Results from this study showed that ALMAS based on particle filtering are relatively robust and accurate for automatic lumbar motion analysis.