Automatic Tracking of Cervical Vertebrae Using Fluoroscopic Sequences

Abstract

The cervical spine has the greatest contribution to the motion of the human skull during full spine flexion and extension. This contribution of individual vertebrae provides precise and valuable information from a diagnosis perspective for patient with neck pains. Previous research related to cervical vertebrae shows that, abnormalities in the cervical vertebrae structures may affect the movement of the cervical spine. To examine the complex structure of cervical vertebrae, doctors and physicians analyze the movement of the vertebrae to diagnose abnormalities throughout the last sequence of the fluoroscopy. Recursively analysis of orientation of each vertebra itself a hectic and sensitive issues for the doctors and physician to target the desired issue. Our study assist doctors and physicians by automate the motion analysis of the annotated data using the kinematic parameters of each vertebrae from diagnosis perspective of neck pains. Using motion estimation of each vertebrae, doctors might be able to find out the issues across the vertebrae and how they are vary throughout last sequence of the fluoroscopy while Flexion-Extension movement. For the validation of motion estimation of annotated data, we have compared our results with the manual selection that is used with the collaboration of the expert doctors and physician. Extraction of motion estimation of each vertebrae, corners of the vertebrae were manually selected at each frame and find out the orientation behavior while Flexion-Extension movement, also compared our results with the other existing algorithms Mean Shift, Kalman Filter, Histogram based matching and Normalized Cross Correlation (NCC). Automatic tracking approach is proposed for the measurement of the cervical spine using Kanade-Lucas-Tomasi (KLT) feature tracking algorithm through fluoroscopic sequences. Physicians and doctor’s intervention will be required at the initial phase just to identify the correct landmarks of the target vertebrae that kinematics parameters are required to obtained, considered them as reference position and extracted translation and rotation of these landmarks in the successive frames using Harris corner detector and use link motion vector for trajectory. This study proposed an efficient solution to track the cervical vertebrae using fluoroscopy sequences and reduce the effort of the doctor by automate the motion estimation which assist them from diagnosis perspective.