Semiautomated image registration for digital subtraction radiography

Objective. The purpose of this study was to evaluate the semiautomatic alignment and correction of affine geometric discrepancies for digital subtraction radiography. Study design. Algorithms were tested in vitro to determine their ability to semiautomatically select reference points on a second image based on points selected on a first (reference) image. A preserved human mandible was imaged with and without bone-equivalent material chips at varying degrees of angulation. Each chip had a mass of less than 10 mg and was no more than 0.3 mm thick. High levels of specificity and sensitivity for chip detection were achieved with 6 degrees of angular discrepancy or less. The algorithms were then applied to radiographs from six human subjects through use of the bone-chip validation model. Results. Sensitivity was 89% and 100% for the three-point and four-point affine warp algorithms, respectively. Specificity for both algorithms was 100%. Conclusions. The data indicate that semiautomated alignment algorithms may enhance the efficacy of digital subtraction radiography while maintaining diagnostic efficacy in clinical trials.