A vision-based surgical instruments classification system

This paper presents a real-time and automatic online vision-based surgical instruments recognition system, which can be used for surgical instruments monitoring during surgery or robotic applications. The main processes of this system consist of feature extraction and classification. In feature extraction, the image of surgical instruments placed on surgical drape are segmented by using color information. Several shape and contour information of the instruments are extracted as features. A two-stage classification scheme based on naïve Bayesian classifier is then proposed to recognize the surgical instruments according to these features. Experimental results demonstrate that the proposed classification scheme can achieve 90.82% accuracy for classifying 7 instruments.