The Navigation System aims to facilitate Minimally Invasive Esophagectomy by intraoperative real-time information about the exact localization of instruments in relation to tumour and lymph nodes. The Navigation System has high accuracy in a static environment and was tested for accuracy and different sources of error in an animal model with similar organ size to humans. The System with Optical Tracking, an immobilization device, preoperative CT-Imaging and Navigation software in MITK on a PC was tested on targets in different esophageal levels in a porcine model. The mean FRE and TRE were 1,75+/-0,83 and 7,4+/-3,2 mm. The Navigation System had lower accuracy with errors caused by different sources of soft tissue deformation that need to be compensated for.
Initial landmark based patient registration with Optical tracking (PolarisTM, NDI) is combined with intraoperative tracking of optical markers on a self-developed navigation instrument. The patient is fixed in a vacuum mattress on a stretcher with additional optical markers to minimize repositioning error. Gastroscopic hemoclips were used as targets(n=14) in the middle and lower thoracic, abdominal esophagus and at the gastroesophageal junction (GEJ) for the accuracy evaluation of the navigation system in a porcine model (n=4). High resolution CT-Imaging of the pig was imported into the navigation system on a conventional PC. CT-data segmentation and target definition was done with MITK (Medical Imaging Interaction Toolkit [4]). After the initial registration the Fiducial Registration Error (FRE) was measured. Laparoscopic preparation of the esophageal hiatus with the pig in the 25° Anti-Trendelenburg-position was then performed. The navigated instrument tip was positioned at each target clip under fluoroscopy control. Target Registration Error (TRE) was measured in the navigation system as off-set from the target to the tip of the navigated instrument.
During the operation the instrument tip was successfully visualized in real-time in relation to the segmented organs by the navigation system at all times. In the accuracy evaluation the mean FRE was 1,75+/-0,83 mm standard deviation. The mean TRE in the accuracy evaluation was 7,4+/- 3,2 mm standard deviation. The error in the X-/Y- and Z-axis differed between the middle and lower thoracic, the abdominal esophagus and the gastroesophageal junction (GEJ) (see Table 1).
In a live animal model with a foregut size that is comparable to human patients the navigation system obtains a mean error of less than 1 cm in a minimally invasive approach to esophagectomy. We believe that the obtained accuracy can help improve intraoperative orientation and identification of lymph nodes and adequate resection margins. The difference between the FRE and the considerably higher TRE can mostly be explained by displacement of the esophagus due to intraoperative iatrogenic manipulation, breathing and cardiac motion. The next step for improved accuracy of the navigation system is the implementation of compensation methods for breathing motion and iatrogenic manipulation. These methods are tested at our institution. Further studies need to approve the system prior to the use with patients. 5