Validation of an automated shape-matching algorithm for biplane radiographic spine osteokinematics and radiostereometric analysis error quantification.

• Main Authors: Craig C Kage, Mohsen Akbari-Shandiz, Mary H Foltz, Rebekah L Lawrence, Taycia L Brandon, Nathaniel E Helwig, Arin M Ellingson
• Format: Article
• Language: English
• Published: Public Library of Science (PLoS) 2020-01-01
• Series: PLoS ONE
• Online Access: https://doi.org/10.1371/journal.pone.0228594

Biplane radiography and associated shape-matching provides non-invasive, dynamic, 3D osteo- and arthrokinematic analysis. Due to the complexity of data acquisition, each system should be validated for the anatomy of interest. The purpose of this study was to assess our system's acquisition methods and validate a custom, automated 2D/3D shape-matching algorithm relative to radiostereometric analysis (RSA) for the cervical and lumbar spine. Additionally, two sources of RSA error were examined via a Monte Carlo simulation: 1) static bead centroid identification and 2) dynamic bead tracking error. Tantalum beads were implanted into a cadaver for RSA and cervical and lumbar spine flexion and lateral bending were passively simulated. A bead centroid identification reliability analysis was performed and a vertebral validation block was used to determine bead tracking accuracy. Our system's overall root mean square error (RMSE) for the cervical spine ranged between 0.21-0.49mm and 0.42-1.80° and the lumbar spine ranged between 0.35-1.17mm and 0.49-1.06°. The RMSE associated with RSA ranged between 0.14-0.69mm and 0.96-2.33° for bead centroid identification and 0.25-1.19mm and 1.69-4.06° for dynamic bead tracking. The results of this study demonstrate our system's ability to accurately quantify segmental spine motion. Additionally, RSA errors should be considered when interpreting biplane validation results.