Image-guided minimally invasive endopancreatic surgery using a computer-assisted navigation system

Background Minimally invasive endopancreatic surgery (EPS), performing a pancreatic resection from inside the pancreatic duct, has been proposed as an experimental alternative to duodenum-preserving pancreatic head resection in benign diseases such as chronic pancreatitis, but is complicated by difficult spatial orientation when trying to reach structures of interest. This study assessed the feasibility and potential benefits of image-guided EPS using a computer-assisted navigation system in artificial pancreas silicon model. Methods A surgical navigation system displayed a 3D reconstruction of the original computed tomography (CT) scan and the endoscope in relation to a selected target structure. In a first step, different surface landmark (LM)-based and intraparenchymal LM-based approaches for image-to-physical space registration were evaluated. The accuracy of registration was measured as fiducial registration error (FRE). Subsequently, intrapancreatic lesions (n = 8) that were visible on preoperative imaging, but not on the endoscopic view, were targeted with a computer-assisted, image-guided endopancreatic resection technique in pancreas silicon models. After each experiment, a CT scan was obtained for measurement of the shortest distance from the resection cavity to the centre of the lesion. Results Intraparenchymal LM registration [FRE 2.24 mm (1.40-2.85)] was more accurate than surface LM registration [FRE 3.46 mm (2.25-4.85); p = 0.035], but not more accurate than combined registration of intraparenchymal and surface LM [FRE 2.46 mm (1.60-3.35); p = 0.052]. Using image-guided EPS, six of seven lesions were successfully targeted. The median distance from the resection cavity to the centre of the lesion on CT was 1.52 mm (0-2.4). In one pancreas, a lesion could not be resected due to the fragility of the pancreas model. Conclusion Image-guided minimally invasive EPS using a computer-assisted navigation system enabled successful targeting of pancreatic lesions that were invisible on the endoscopic image, but detectable on preoperative imaging. In the clinical setting, this tool could facilitate complex minimally invasive and robotic pancreatic procedures.