Performance and Robustness of Automatic Fluoroscopic Image Calibration in a New Computer Assisted Surgery System

In order to improve the clinical usefulness of computer-assisted fluoroscopic navigation, a new algorithm to automatically determine the calibration of fluoroscopic images has been developed. This is a challenging task since the intraoperative images acquired from fluoroscopic systems are often poor, making detection of the calibration grid difficult. Several feature-based methods have been implemented to perform bead detection for automatic detection of the calibration grids. The algorithms include support for multiple fields of view, a feature not supported on any computer assisted systems to date. In order to evaluate the performance of the algorithms, special phantoms were made and a cadaver study was performed to challenge the algorithms. One hundred images were acquired using three different C-Arms (OEC 9600, OEC 9800 and Philips BV-300+) using two different fields of view (nine and twelve inch). The chosen method successfully registered the images in ninety-six of the cases. The images that were not successfully registered were of limited clinical value anyway due to the very poor image quality.