The Learning Curve of Registration in Navigated Skull Base Surgery

Background: Computer aided surgery (CAS) is advantageous in challenging procedures in head and neck surgery. It is not clear, if the application of CAS has to be trained to achieve reliable results. The learning curve of the registration of the patient's coordinates to prior acquired radiologic imagery was investigated. Material and methods: 4 residents performed pair-point registrations on 5 anatomic specimens in an experimental wet lab. The residents were in the same year of education and had no experience in CAS procedures. After each registration the application error was evaluated by determining the target registration error (TRE). Pair point matching by skin glued external fiducials was compared with pair point matching by internal anatomical landmarks. Results: The application accuracy was improved by increasing numbers of performed registrations (p < 0.001, trendtest of Page). An inverse trend of the learning curve could be observed, the median TRE values improved from 3.3 mm in the first registration to 1.6 mm after the fifth registration. In comparison e.g. an experienced CAS-user can achieve submillimetric TRE values under wet lab conditions. Pair point matching by anatomical landmarks resulted in worse application accuracy initially and the learning curve was steeper than with external fiducial markers. Conclusion: There is a training effect in CAS interventions. Pair point matching results in sufficient application accuracy after training only.