Studies on Acquisition of Training Data in Endoscope Training System with Surgery Video

-To use the surgery video as training data-

Abstract

In this study, we have developed an endoscope training system that uses training data from a qualified surgeon of forceps operations. The system has received positive evaluations from doctors as a method for helping train beginners. However, the system has the problem that the training data is limited. Thus this study aims to expand the training data by estimating the position information of the forceps. The constructed algorithm estimates the position information of the forceps in the videos used as part of the training data along the time series using machine learning, evolutionary computation, and computer graphics. This paper describes the technical elements of the constructed system and reports on the verification results of the accuracy of the estimated forceps position information.