Design of a Head Position and Orientation Detection System Based on Improved Markerless Visual Tracking for Unrestrained Awake Mice

During nuclear medicine imaging of unrestrained awake mice, head movement can lead to motion artifacts that compromise image resolution and alter tracer distribution. Consequently, eliminating head motion artifacts is crucial for improving the quality of brain positron emission tomography (PET) images. In this study, we designed a head position and orientation detection system based on improved markerless visual tracking for unrestrained awake mice. The system comprised five primary phases: image preprocessing, image segmentation, three-dimensional (3D) reconstruction, construction of the complete reference point cloud, and point cloud registration. To evaluate the performance of the designed motion tracking system, repeatable measurements of the head position and orientation of a stationary mouse phantom were conducted. Experimental results showed that the root mean square errors (RMSE) for the translational degrees of freedom were 0.051 mm, 0.222 mm, and 0.243 mm, and the RMSE for the rotational degrees of freedom were 0.035°, 0.077°, and 0.072°. Statistical analysis of the repeatable measurement results of the translational and rotational parameters by Quantile-Quantile Plots and the Kolmogorov-Smirnov test confirmed that each parameter followed a normal distribution. The designed motion tracking system accurately measured the head position and orientation during nuclear medicine imaging, and statistical analysis further demonstrated the system’s reliability. Additionally, the system did not rely on markers, avoiding the risk of marker slippage and detachment. The motion information obtained from this system could be utilized to compensate for head motion artifacts and enhance the quality of nuclear medicine images.