Stabilization of Particle Filter Based Vertebrae Tracking in Lumbar Spinal Videofluoroscopy

Abstract

Tracking of spinal vertebrae in videofluoroscopy video is useful for diagnosis of certain spine pathologies, such as vertebral fractures, spondylolisthesis and low back pain. The aim of this work is to improve a semi-automatic method for vertebrae tracking based on particle filter. The process starts with vertebrae model construction by splining hand-selected landmark points, then the particle filter tracks the vertebrae in each frame. While the process successfully tracks the vertebras, the trajectories are quite jittery This paper aims to improve the overall tracking by smoothing the tracking trajectory using curve fitting function. The tracking results are more natural with very little noise.