Research on Real-time Surgical Navigation System Based on Machine Learning and Fusion of Multimodal Sensors

Abstract In response to the significant challenge of occlusion in optical navigation, this paper proposes an innovative surgical navigation instrument positioning technique based on the fusion of Ultra-Wideband (UWB) and Inertial Measurement Unit (IMU) data. By installing UWB positioning modules and IMU sensors on the surgical instrument arm, this study introduces a machine learning-based data fusion method to fully utilize UWB ranging and IMU high-precision roll and pitch angles, addressing the issues of attitude angles and Z-axis accuracy in UWB pose estimation. The method employs machine learning techniques to model the complex nonlinear relationships among three sets of UWB ranging values, two IMU attitude angles, and the position and yaw angle of the surgical instrument. Experimental results demonstrate that the proposed method achieves an average absolute error of less than 2cm in three-axis position and less than 1° in three-axis attitude angles. This surgical navigation instrument positioning technique exhibits significant advantages in realizing an efficient and accurate surgical navigation system. By optimizing the positioning accuracy of surgical instruments, this technology offers a novel approach to enhancing the performance of surgical navigation systems, providing innovative solutions for positioning requirements in medical practice.