Automatic Detection of Cervical Vertebral Landmarks for Fluoroscopic Joint Motion Analysis

Real-time motion assessment of the cervical spine provides an understanding of its mechanics and reveals abnormalities in its motion patterns. In this paper we propose a vertebral segmentation approach to automatically identify the vertebral landmarks for cervical joint motion analysis using videofluoroscopy. Our method matches a template to the vertebral bodies, identified using two parallel segmentation approaches, and validates the results through comparison to manually annotated landmarks. The algorithm identified the vertebral corners with an average detection error under five pixels in the C3–C6 vertebrae, with the lowest average error of 1.65 pixels in C4. C7 yielded the largest average error of 6.15 pixels. No significant difference was observed between the intervertebral angles computed using the manually annotated and automatically detected landmarks ($$p>0.05$$). The proposed method does not require large amounts of data for training, eliminates the necessity for manual annotations, and allows for real-time intervertebral motion analysis of the cervical spine.