Action Recognition of Interventional Surgery Based on 3D Convolutional Neural Network

• 1、School of Artificial Intelligence, Beijing University of Posts and Telecommunications

Abstract：Vascular disease is a disease with a very high fatality rate, which seriously threatens the life and health of patients. Interventional therapy is a minimally invasive, high-tech treatment technique with the characteristics of less trauma, quick postoperative recovery, more indications, and fewer complications. The interventional doctor\'s action recognition is the basis of the operation stage analysis, intraoperative risk warning, and operation evaluation system. Most of the current research on surgical motion recognition requires doctors to wear sensors. In this paper, computer vision technology is used without wearing any sensors, and the input video is analyzed and processed through a convolutional neural network to realize the recognition of interventional surgical actions. This paper proposes a two-stage network recognition architecture to improve the robustness and efficiency of surgical action recognition. A lightweight 3D CNN is used for surgical action detection and a deep 3D CNN is used for surgical action classification. When the detector network detects an action, the classifier network starts to work, reducing the power consumption of the system. Both the detection network and the classification network have good performance and are suitable for clinical deployment.

Keywords： Interventional surgery Action recognition Computer vision 3D convolutional neural network