gms | German Medical Science

Introduction: Functional endoscopic sinus surgery (FESS) is a frequently carried out ENT-intervention. The patients risk results from the complexity of the procedure and anatomical proximity of orbit and skull base to the operation area. Decreasing availability of human compounds and a tight OR-schedule require new concepts in surgical training of young ENT-surgeons. We developed a virtual training system for FESS that allows tissue interaction providing force feedback.

Material & Methods: The virtual model for endoscopic sinus surgery was generated using a CT-dataset of a real patient and the software tools VESUV©, KisMo© and KISMET. First, this dataset was segmented semiautomatically in VESUV©. In KisMo© we generated a surface model of the right nasal cavity and modelled the lower and middle turbinate and a few ethmoid air cells. The used instruments (endoscope, forceps) were modelled using the 3 D visualisation and modelling software Rhinoceros©. The modelled structures were connected and texturized using original images and video material taken of the intranasal anatomy during an operation. The scene was implemented in the simulation software KISMET and connected with the force feedback device PHANToM Premium (SensAble Technologies Inc., USA).

For evaluation of the system, a pilot test series was started. Two test groups with ten test subjects in each group had to perform sinus surgery. The first group consisted of experienced ENT-surgeons, the second group of endoscopy inexperienced medical students. The test subjects had to perform in two sessions five procedures per session with a one week interval between session one and session two. The training data was recorded by an integrated software tool (KISMET Tutor) that logged start and end time, error rate and error type during the procedure. For evaluation of the data mean procedure time and mean error rate were calculated and compared between both groups.

In a follow-up study with 4 inexperienced test subjects that performed the procedure 50 times (ten session with every five procedures). Additionally, a questionnaire had to be filled by the test subjects.

Results & Conclusion: Evaluation of the data of the pilot test series showed no significant learning curve concerning time and error rate for the inexperienced surgeons. The data of the follow-up-study also supplied no statistically significant learning curves but revealed a trend to less time and error rate by training the procedure. (Regarding the data leads to the assumption that the number of training sessions of the pilot test series is too little to record a learning curve.) Another aggravating point is the insufficiency of the haptic properties, it is a balancing act between real time rendering and realistic virtual training environment.

The questionnaire showed a high level of realistic anatomy, texturing and visualisation. The analysis of the scene and the resulting learning curves suggests that an effective simulation system has to provide a certain closeness to reality to result in statistically significant learning curves. Correct haptic properties can be supportive.