Evaluation of a semiautomatic 3D fusion technique applied to molecular imaging and MRI brain/frame volume data sets

A generally applicable 3D fusion method was evaluated using molecular imaging and MRI volumetric data sets from 15 brain tumor patients with stereotactic frames attached to their skull. Point pairs, placed on the frame only, were chosen, polynomial warping coefficients were generated to map voxels from one coordinate space to the other. The MRI frame was considered the reference structure and the standard for "correct" registration. An ANOVA test (p > 0.05) confirmed the point pair choice to be consistent. The 95% confidence interval for the t-test showed the measured distance difference between the registered volumes was within one MRI voxel. A further experiment was conducted to independently evaluate the brain registration based on testing for consistency of randomly selected interior/exterior points. A t-test result (p < 0.05) showed that the consistency (i.e., both interior or both exterior) before and after volume registration were significantly different. This fusion method may be a viable alternative when other methods fail.