Color Based Segmentation in Monocular System for Prosthesis Pose Estimation During Total Hip Replacement Surgery

Shaojie Su,Guangli Jiang,Hong Chen,Zhihua Wang
DOI: https://doi.org/10.1109/biocas.2016.7833796
2016-01-01
Abstract:Safe-zone in total hip replacements (THR) means that the acetabular cup is placed within 30 to 50° of abduction and 5 to 25° of anteversion. To help surgeons with prosthesis placement into the safe zone and real-time prosthesis pose estimation, a femoral head prosthesis with a camera mounted inside is designed. To ensure the validity of this monocular vision based system in blood covered situation, an RGB color based segmentation method is proposed in this paper. Customized patterns are printed on the inner surface of acetabular cup. When put into the acetabular cup, the femoral head prosthesis will keep collecting images of the customized patterns. Relative pose between the prostheses is then calculated by matching feature points between frames. Since blood is red in color, we extract red channel of the image sequence for processing to ensure high contrast between the patterns and background. Top-hat preprocessing is also applied to eliminate influence of uneven illumination. Experiment results show the mean error of the monocular vision based pose estimation is 1.426°, which is within tolerance compared to the range of safe zone. The RGB color based segmentation achieves a pattern recognition rate as high as 94.1% in blood covered situation, which is 4 times of that of ordinary method.
What problem does this paper attempt to address?