Enhanced CAS in head and neck-surgery: evaluation of intraoperative data modality iMRI and sonography

Abstract

Purpose: Computer-assisted surgical procedures are well established in head/neck- and neurosurgery. Most systems are based on preoperative CT or MR data. To visualize the surgical change and re-registration, there is a need for intraoperative imaging. Currently, only few imaging methods are available to supply up-to-date information on the surgical progress, e.g., inter-/intraoperative MRI, interoperative CT, and sonography/ultrasound. The purpose of this study is to compare procedures and algorithms for intraoperative co-registration of iMRI and sonography in head/neck- and neurosurgery. Methods: Intraoperative imaging was performed by interventional MR (GE Signa SP) and high-resolution ultrasound (Siemens Sienna). iMRI: Localite Navigator Software (enhanced CAS) allows integration of preoperative data to intraoperative imaging of MR, fMR and CT. The co-registration was based on fiducial markers. The implementation was designed as a multimode real-time navigation system supplying images of bony structures/contours (derived from CT data) fused into real-time MR images. A total of 10 markers (three forehead, three temporal scalp, three mastoid, and one occipital scalp) were used in this study to define reproducible fixation points. Sonography: Based on neuronal networks, a recognize-algorithm for co-registration was developed using sonography data sets of the neck. Results: The statistical point-to-point accuracy (marker difference) of co-registration in the open MR was determined to be 1.2±0.1 mm. The co-registration CT preop and iMRI are reliable and nowadays established in selected surgical procedures of the skull base, e.g., biopsy of the petroclival region. The system yields a significantly better image quality and higher repetition rate than the standard real-time images. Typical for any marker-based registration, the matching accuracy for the registered data sets depends on the accuracy of locating the markers in the respective images, and the inherent displacement of the fiducial markers. The markerless co-registration Sono preop/Sono intraop based on neuronal networks is still less reliable with a positive recognition rate of 46/51 of structures with a maximal volume of 2.110 cm 3. Recognized structures can be used for co-registration with different data modalities, e.g., MR, CT or for single-modality navigation. Conclusions: Both interventional MRI as well as sonography allows intraoperative data update. The marker-based co-registration CT preop/iMRI is still more reliable than Sono preop/Sono intraop based on neuronal networks. Further developments tend towards the improvement of mutual information algorithm and marker-like landmarks to improve the recognition rate of integrated sonography. Low costs and less surgical tools expenditure are the advantages of sonography, and it will be a serious possibility for intraoperative CAS-update. Otherwise, the study demonstrates the possibility of combining intraoperative MR and sonography with other modalities for enhanced CAS in ORL and neurosurgery.