gms | German Medical Science

Objective: DT images usually show a certain degree of distortion in comparison to the standard anatomic T1 MR images. Therefore, DTI based data such as tracked fibers may be associated with a certain degree of inaccuracy when projected on T1 images which are used for navigation. Elastic image fusion is a newly available method to correct for this distortion. The present study aims to evaluate the accuracy of an automatic elastic fusion algorithm regarding the fusion of DTI to T1 images.

Method: 45 anatomical landmarks were marked in preoperative MR image pairs where geometric distortion was to be expected (non-distorted T1 with contrast enhancement and the first EPI-distorted B0 image) in a sample of 13 patients expected to undergo tumor resection.

Images were then processed with a prototype of the Brainlab Elastic Image Fusion software and landmarks in the distorted DTI image were distortion-corrected through elastic deformation. To quantify the displacement, the residual error, the Euclidian distance of the deformed landmarks to the landmark position resulting from a rigid image fusion were contrasted with the residual Euclidian distance of the deformed landmarks to their true position in the image. The results were further analyzed with respect to the distance of the landmark to the center of the tumor in order to investigate the impact of local anatomical abnormalities on the precision of the algorithm.

Results: 617 landmarks were set, 51 of them had to be excluded because of imprecise placement (rigid distance over 2cm).

There is a significant correlation (r=-0.47, p<0.0001) between improvement/detoriation after elastic fusion and the amount of deformation of landmarks. This means the more a point is moved by the algorithm, the higher is the distortion correction. 30% had been moved more than 3mm. In the group with 30 slices we saw improvement of 10/13 (77%) points and 3/13 (23%) worsened. In the group with 73 slices 143 /233 (61%) improved and 90/ 233 (39%) were worse.

Conclusions: Elastic fusion is a method for correcting for image distortion. Our present data show that elastic fusion in its present form increases fusion accuracy, however, not to the same extent for each region. While the present evaluation assessed correction for image distortion, the major use for elastic fusion would be the fusion of post resection images to pre-resection images using different imaging modalities to evaluate i.e. extent of tumor resection by intraoperative CT instead of MRI.