EVALUATION OF IMAGING GEOMETRIES CALCULATED FROM BIPLANE IMAGES

A technique is developed that will calculate accurate and reliable ima ging geometries and three-dimensional (3D) positions from biplane imag es of a calibration phantom. The calculated data provided by our techn ique will facilitate accurate 3D analysis in various clinical applicat ions. Biplane images of a Lucite cube containing lead beads 1 mm in di ameter were acquired. After identifying corresponding beads in both im ages and calculating their image positions, the 3D positions of the be ads relative to each focal spot were determined. From these data, the transformation relating the 3D configurations were calculated to give the imaging geometry relating the biplane views. The 3D positions of o bjects were determined from the biplane images along with the correspo nding imaging geometries. In addition, methods are developed to evalua te the quality of the calculated results on a case-by-case basis in th e clinical setting. Methods are presented for evaluating the reproduci bility of the calculated geometries and 3D positions, the accuracy of calculated object sizes, and the effects of errors due to time jitter, variation in user-indication, centering, and distortions on the calcu lated geometries and 3D reconstructions. The precision of the translat ion vectors and rotation matrices of the calculated geometries were wi thin 1% and 1 degrees, respectively, in phantom studies, with estimate d accuracies of approximately 0.5% and 0.4 degrees, respectively, in s imulation studies. The precisions of the absolute 3D positions and ori entations of the calculated 3D reconstructions were approximately 2 mm and 0.5 degrees, respectively, in phantom studies, with estimated acc uracies of approximately 1.5 mm and 0.4 degrees, respectively, in simu lation studies. This technique will provide accurate and precise imagi ng geometries as well as 3D positions from biplane images, thereby fac ilitating 3D analysis in various clinical applications. We believe tha t the study presented here is unique in that it represents the first s teps toward understanding and evaluating the reliability of these 3D c alculations in the clinical situation. (C) 1998 American Association o f Physicists in Medicine. [S0094-2405(98)01606-X].