Effect of the number of external markers in multimodal 3-D image registration

Abstract

The minimum number of external markers that is needed for a multimodal 3-D image registration is studied quantitatively by using Monte-Carlo techniques. Using computer simulation, test images are generated over the head in both modalities using a predefined transformation. The optimized transformation obtained from a set of matched external markers using the simplex minimization method followed by a variable metric method is applied to these images to study how good the registration is with a given number of markers. The authors studied the /spl chi//sup 2/ per degree of freedom and the residuals as a function of the number of external markers. It is found that the registration is optimum for 10 markers and above.<>