An Artificial Neural Network approach for Haptic Discrimination in Minimally Invasive Surgery

In this paper we investigate the possibility of processing the tactile perception by using a novel biomimetic approach for the pattern recognition module. The goal is to enhance the perception in complex virtual environments deriving from haptic displays mimicking human tactile discrimination. To do this we explored a Minimally Invasive Surgery application where the tactile information are strictly limited. In fact, this promising technique suffers from some evident limitations due to the surgeon loss of tactile perception during palpation of internal organs. This is basically due to the mechanical transmission of the elongated tools used during operation. We propose to integrate an Artificial Neural Network in an electronic board capable of processing data provided by a sensorized laparoscopic tool. The capabilities of several pattern recognition techniques present in literature, the Principal Component Analysis (PCA), a Multilayer Perceptron (MLP) and a Kohonen Self-Organising Map (KSOM) are investigated. The results are compared with that obtained psychophysically on five viscoelastic materials.