融合实时超声影像的多模态肌肉运动特性研究

摘要

人体肌肉活动是一个复杂过程,常用的研究方法包括肌电信号、肌音信号、超声影像等.提出一种多模态方法,联合这三种信号对肌肉活动特性进行研究.研发了多通道运动信号采集系统,对包括超声影像在内的多种信号进行实时同步采集.开发了一种新的图像追踪算法,实现肌肉边界的自动追踪.对膝关节伸展时股直肌等长收缩的情况进行研究,对采集的各种信号进行特性分析,并联合多种参数对关节力矩进行估计.研究表明,这三种 信号的线性组合与关节力矩之间存在显著的线性关系,拟合度达到0.99,高于分别用3种信号进行线性拟合的情况.与单独研究一种信号相比,这种多模态的方法能提供更多关于肌肉运动的信息,为进一步深入研究提供一种可行的方法.%The in vivo behavior of human muscle during contraction is a complicated process, and some commonly-used research techniques include electromyography ( EMG) , mechanomyography ( MMG) , and ultrasonography (US). This study proposed a new multi-model method which integrated EMG, MMG and US together to investigate the muscle activities. An acquisition system was developed to simultaneously collect US images and other signals in real-time; and an image tracking algorithm was implemented to automatically extract the muscle boundary in US. This study examined the responses of human rectus femoris muscle during voluntary isometric ramp contraction for kriee extension. The collected signals were analyzed and three parameters extracted from EMG, MMG and US were combined to estimate the torque. The results show that there are significant multiple linear relationships between the three parameters and torque. The R of the regression was 0. 99 which was much better than the linear regressions with single parameter. In conclusion, this multi-model method can provide more information about the muscle activities, and further research will be conducted for better understanding of the muscle contraction mechanism.