EXPLORE PUBLICATIONS BY COUNTRIES

Instrument recognition , Image analysis, SIFT transform, Support Vector Machines

In this work, an approach for surgical instruments recognition within images taken from an endoscope video camera is presented. This approach is based on the analysis of the images using the SIFT transform and a clustering method called k-means, jointly with the use of Support Vector Machines. The instrument identification might be used for recognizing the action the surgeon is performing during the intervention, which may be useful for monitoring or training purposes. By correlating the action which is being performed with the intervention protocol, a robotic assistant might warns about the correct order in which these actions should be done, or the time they should take, according to average measurements normally accepted by medical organizations. Other approaches, based on the analysis of the instruments trajectories and forces/torques exerted by the surgeon have been proponed by Rosen et al. The That approach implies the need for attaching sensors to the handles used by the surgeon to manipulate the laparoscopic tools, which not always is possible or might be not admitted by the surgeon for his own comfort ability. The original aspect of this work is to take the images directly from the Camera Control Unit (CCU) of a laparoscopic system, which provides video signal output of the embedded camera, instead of wiring additional sensors, which have to be connected and calibrated each time a new intervention is started.