论文部分内容阅读
目的:设计一种基于标记点的实时动态耳廓三维影像导板，并对其进行评价。方法:以1个商品化人头模型以及2019年8月于中国医学科学院整形外科医院招募的3名志愿者为研究对象，其中男性2名，女性1名，年龄24～27岁。运用ITK-SNAP 3.6.0和MeshLab软件(V 2016.12.23) 对3名志愿者的头颅CT图像进行三维重建，获得虚拟耳廓三维模型。通过人头模型论证标记点放置的最佳数量为3个，最佳位置为眉角、鼻尖及耳垂下点，利用这3个标记点将虚拟耳廓三维模型与对应的真实耳廓进行关联注册，通过投影仪将虚拟耳廓三维影像投射至对应的真实耳廓，红外线动态捕捉仪捕捉标记点位置变化对投射图像进行实时跟踪，通过计算注册后的重合误差率、跟踪过程中的重合误差率、投射图像变形率和追踪延迟度，评价实时动态耳廓三维影像导板的注册精准度、跟踪精准度、投射图像变形度及跟踪延迟度。结果:注册后平均重合误差率为2.5%(正常人双侧耳廓大小差距为2.7%)；在顺时针旋转30°内跟踪平均重合误差率小于2.7%；在顺时针旋转30°与逆时针15°范围内投射图像变形率小于2.6%，一定旋转角度范围内平均重合误差率和投射图像变形率小于2.7%，说明一定旋转角度范围内跟踪精准度高，投射图像变形度小，跟踪延迟时间均小于0.1 s，满足手术需求。结论:成功构建了基于标记点的实时动态耳廓三维影像导板，其注册精准度高、跟踪精准度高、投射图像变形度小、跟踪速度快，为将来其应用于临床耳廓再造术打下基础。“,”Objective:To design a real time dynamic three-dimensional(3D) auricular image guidance based on sensors and make an evaluation of it.Methods:One commercial head model and 3 volunteers (2 men and 1 woman, aged 24-27 years old) recruited from Plastic Surgery Hospital, Chinese Academy of Medical Sciences in August 2019 were studied. Digital models of auricle were made using CT data of three volunteers with ITK-SNAP 3.6.0 and MeshLab (V 2016.12.23). The optimal number of sensors was three and the optimal location of sensors was one side of the eyebrow, the nasal tip and the inferior point of the earlobe. The virtual model was connected with the real auricle based on three sensors. The 3D auricular image was projected to the corresponding real auricle through the projector. The infrared motion-capture device was used to capture the changes of the position of the sensors to realize real-time tracking. The matching accuracy between 3D auricular image and real auricle was evaluated by calculating the coincidence error rate after registration, calculating the coincidence error rate, image deformation and tracking latency during tracking.Results:The average coincidence error rate after registration was 2.5% (the difference between the normal bilateral ears was 2.7%), which showed the registration precision was high. The coincidence error rate was lower than 2.7% in certain rotation angle (30° clockwise) and the image deformation was lower than 2.6% in certain rotation angle (30° clockwise and 15° counterclockwise) during tracking, which demonstrated the precision of tracking head movement was high and image deformation was low in a certain rotation range. The time of image update and motion tracking was less than 0.1 s, which satisfied the requirement of surgery.Conclusions:This study demonstrates that the fabrication of real-time dynamic 3D auricular image guidance based on sensors is feasible and the image guidance has high accuracy of registration and tracking, low image deformation and fast tracking speed, which lays a foundation for its application in clinical auricle reconstruction in the future.