摘要
以粒子滤波理论为基础，计算、分析腰椎椎体运动轨迹及其物理参数。以模具腰椎椎体为研究对象，拍摄10组连续的数字荧光视频影像，每组重复2次做屈伸运动，并用固定在各椎体中心的刚性笔实时记录实际轨迹。对于算法实现测得的值与实际测量的值做相关分析，显示出强相关。测试算法的变异性和鲁棒性，计算5块椎体的旋转角度和椎体中心位移的标准离差（RMS）与测量标准误差（SEM）。X位移方向：RMS分布于0．25～0．93mm，Y位移方向：0．47～0．87mm，矢状屈伸旋转角度：0．96°～1．40°（P〈0．05）。重测信度分析测试中，5块椎体的角度、位移参数的对应相关系数均大于0．9，呈强相关。根据序贯重要性重采样粒子滤波算法原理，实现了模具腰椎在矢状弯曲运动过程中的运动轨迹检测。通过比较分析算法测得和实际测得参数的相关性、算法本身的变异性和鲁棒性及重测信度，我们认为在一定精度内，该算法稳定性好，一致性高。
To calculate trajectory and analyse physical parameters of lumbar vertebrae based on the particle filter. A lumbar vertebrae model was enrolled into this study. Digitised video fluoroscopy was obtained during sagittal flexion and extension. Each vertebra trajectory was recorded by real - time depiction of the vertebral body with rigid fixation pens when the collection was finished. Continuous dynamic lumbar inter - vertebral flexion - extension was assessed by digitized video fluoroscopy in calibration model at 10 times. Each time included 2 integral cycles, ln the 20 integral cycles, correlation coefficient of x- and y- translation were calculated correspondingly between the automatic tracking measurement and actual measurement, which was strong related. For each integral sequence, root mean square differences （RMS） and standard error of the measurement （SEM） of rotation angle as well as the x - and y - translation of the centre among 10 trials were calculated to test the variability and robustness. From the results of RMS and SEM, x - translation was ranged from 0.25 to 0.93 ram. Y - translation was from 0.47 to 0.87 mm. Angle of rotation, from 0.96° to 1.40° （ P 〈 0.05 ）. In the test - retest trails, the CC 〉 0.9 manifested that the measurement was of good stability and of high consistency over time. According to the principle of sequential important resampling particle filter, we implement movement detection of lumbar vertebrae calibration model. The analysis to the variability and robustness of particle filter algorithm itself and correlation coefficient and test - retest reliability show well in stability and consistency between the results computed from algorithm and actual measurement.
出处
《生物医学工程研究》
2009年第2期91-95,99,共6页
Journal Of Biomedical Engineering Research
基金
中国医学科学院北京协和医学院生物医学工程研究所与大庆石油管理局重点科技项目(QR/AD/7.1-1-03)
中央级公益性科研院所科研业务专项资助项目
关键词
粒子滤波
非线性滤波
序贯重要性重采样
脊柱
数字荧光透视
自动跟踪
Particle filter
Non- linear filter
Sequential important resarnpling
Spine
Digitised video fluoroscopy
Auto tracking
