Real-time integration of ultrasound into neuronavigation: technical accuracy using a light-emitting-diode-based navigation system

Summary.Background. In brain surgery, intraoperative brain deformation is the major source of postimaging inaccuracy of neuronavigation. For intraoperative imaging of brain deformation, we developed a platform for the integration of ultrasound imaging into a navigation system.Method. A commercially available ultrasound system was linked to a light-emitting-diode- (LED) based neuronavigation system via rigid fixation of a position localiser to the ultrasound probe and ultrasound image transfer into the navigation system via a S-VHS port. Since the position of the ultrasound image co-ordinate system is not readily defined within the navigation reference co-ordinate system (REF CS), a transformation which links both co-ordinate systems has to be defined by a calibration procedure. Calibration of the ultrasound probe within the REF CS was performed via a cross-wire phantom. The phantom target was defined within the navigation co-ordinate system (by pointer under microscopic control) and imaged by ultrasound. Ultrasound presets were optimised (digital beam focusing, gain intensity) to attain a small echoic target for manual target definition. The transformation was derived from 150 ultrasound measures and iteration. Accuracy was calculated as mean linear error (LE; in XREF, YREF, or ZREF direction), overall mean LE (linear errors of all axes XREF to ZREF) and Euclidean error (EE; vectorial distance from the physical target).Findings. Optimised ultrasound presets (8 MHz frequency, digital beam focusing, 20% gain intensity) enabled a low interobserver error (mean: 0.5 mm, SD: 0.28) for target definition within the 2-D ultrasound image. Mean accuracy of pointer-based physical target definition in the REF CS was 0.7 mm (RMSE; SD: 0.23 mm). For navigated ultrasound, the overall mean LE was 0.43 mm (SD: 1.36 mm; 95%CL: 3.13 mm) with a mean EE of 2.26 mm (SD: 0.97 mm; 95%CL: 4.21 mm).Interpretation. Using a single target cross-wire phantom, a highly accurate integration of ultrasound imaging into neuronavigation was achieved. The phantom accuracy of integration lies within the range of application accuracy of navigation systems and warrants clinical studies.