Maneuvers recognition in laparoscopic surgery: Artificial Neural Network and hidden Markov model approaches

Abstract

The work presented in this paper is focused on movement recognition as a first step to achieve the automation of a two-arm-surgical-robotic-system in the laparoscopic surgical environment. In order to accomplish coordination between the surgeon and the robotic assistant, a system able to recognize and differentiate between certain standard surgical maneuvers should be developed. Two different methodologies are proposed to model and identify several surgical maneuvers. The first method is based on Artificial Neural Networks (ANN), by codifying the movements through their Fourier spectra and the second one is based on HMMs which represents the interaction between the surgical tools. The proposed approaches will be tested through a set of experiments that mimic surgical movements as in tissue cutting, suturing and transporting. In this way, the recognition system is able to distinguish between the different maneuvers which have been modeled.