Toward the design of a novel surgeon-computer interface using image processing of surgical tools in minimally invasive surgery

摘要

Minimally invasive surgery (or key-hole surgery) is an alternative to open surgery has been gaining popularity among patients and health delivery systems. In general, to view the surgical site, an endoscope is inserted into the abdominal cavity though natural or artificial incision. Long stem surgical tools are also inserted through supporting incisions. The surgeon can then perform the operation by indirectly viewing the scene and manipulating the surgical tools. While viewing the monitor, the surgeon do not have any automatic access to preoperative images or patient specific data or be able to manipulate superimpose them on the viewing monitor. This paper presents a novel approach based on image processing of the surgical site and neural network framework for classifying and identifying gestures of surgical tools and classification of their motions. Seven feature quantities were selected as an input to a feed-forward neural network. Experimental analysis of the classification was carried-out for single tools and multiple tool gestures in an in-vitro setting. Through a number of trails we were able to demonstrate the feasibility of our gesture recognition approaches.