AnArtificial Neural Network approach forHaptic Discrimination inMinimally Invasive Surgery

In thispaperwe investigate thepossibility of processing thetactile perception byusinga novelbiomimetic approach forthepattern recognition module. Thegoalisto enhance theperception incomplex virtual environments deriving fromhaptic displays mimicking humantactile discrimination. To dothiswe explored aMinimally Invasive Surgery application wherethetactile information arestrictly limited. Infact, this promising technique suffers fromsomeevident limitations dueto thesurgeon loss oftactile perception during palpation ofinternal organs. Thisisbasically duetothemechanical transmission ofthe elongated tools usedduring operation. We propose tointegrate anArtificial NeuralNetworkinanelectronic boardcapable of processing dataprovided byasensorized laparoscopic tool. Thecapabilities ofseveral pattern recognition techniques present inliterature, thePrincipal Component Analysis (PCA), aMultilayer Perceptron (MLP)andaKohonenSelf-Organising Map (KSOM)areinvestigated. Theresults arecompared with thatobtained psychophysically onfive viscoelastic materials. IndexTerms-Minimally Invasive Surgery, tactile perception, haptic display, artificial neural networks.