Evaluation of accuracy of a novel elastic fusion algorithm — a prospective clinical study

Abstract

Background: Although rigid fusion is generally accepted as gold-standard method, it often stands for time consuming and manual registration procedure. A novel elastic fusion algorithm is introduced to overcome these drawbacks. Objectives: To investigate image registration accuracy of bony structures and adjacent soft tissues provided by novel elastic fusion algorithm. Methods: Five patients (3 male, 2 female, mean of 57 years) with outboard obtained computed tomography (CT) imaging of the head and neck were included in the study. For each, a 3-Tesla (3 T) magnetic resonance imaging (MRI; Philips Healthcare, Best, The Netherlands) of the same region was conducted. The MRI protocol included three-dimensional (3D), axial and sagittal two-dimensional scans showing different contrasts and was repeated three times with the patient in neutral, flexed and rotated head position. To verify accuracy of novel spine curvature correction method (embedded in the ImageFusion Element, Brainlab AG, Feldkirchen, Germany), 80–120 bony landmarks were defined in the CT image, transferred via rigid fusion and used as starting value for elastic fusion algorithm. Finally, Euclidian distances between rigidly and elastically transferred landmarks were determined. Findings: Although the image data revealed strong translational and rotational displacements of vertebrae in both modalities (due to varying head positions), 95% of the landmarks showed an error below 3 mm. However, using 3D MRI data provides highest registration accuracy and co-registration of axial MRI acquisitions is superior than compared to sagittal MR images. Conclusion: Generally, it can be stated that elastic fusion provides high MR-to-CT image registration accuracy and represents a considerable step towards soft tissue registration and navigation.