Intelligent Recognition of Artificial Joint Wear Particles From Lubrication Fluid Based on Multi-View Image Features

Implanted artificial joints will produce a large number of abrasive particles due to friction. These abrasive particles not only aggravate the wear of the friction pair but also have a series of biological reactions with human tissues, which will affect the service life of the joints and the health of patients. Therefore, studying the types and generation mechanism of abrasive particles is of great significance to improve the reliability and service life of artificial joints. However, the traditional artificial joint wear particle analysis methods require complicated operations such as tissue fluid decomposition, dilution, centrifugation, and filtration, which are time-consuming and labor-intensive, and the chemical reagents used can also cause harm to the human body. To improve the automation level of artificial joint wear particle analysis and reduce manual intervention, an artificial joint wear particle analysis method is here proposed. This method is based on using image sequences for rapid extraction and classification of wear particle types. First, moving wear particles in the video are detected and tracked; then, extract the contour features of each particle for single-view recognition; finally, merge multi-view processing and Intelligent recognition to realize quantity statistics and morphological classification of artificial joint wear particles. Compared with the traditional analysis approaches, the proposed method achieves direct and rapid acquisition of the number and types of wear particles from the tissue fluid. This method can significantly reduce the labor and material costs, improve the analysis efficiency, and the wear state of the friction pair of the artificial joint.