Empirical Mode Analysis for Characterization of Hand Tremor in the Design of Laparoscopic Tools1

摘要

Characterization of fatigue induced hand tremor in laparoscopic surgical scenario is important in order to compensate its effect in robot assisted surgery (RAS). While the conventional analysis of muscle fatigue is highly varying in nature, there exists a constant search for fatigue induced hand tremor assessment and compensating device for surgeons to improve the quality of surgery. Moreover, detection of fatigue in dynamic activity itself is challenging. We have tried to address this issue by using a nonparametric method called empirical mode decomposition (EMD) of signals. Instead of analyzing the raw signals, spectral analysis of selective mode seems to be better for detecting the muscle fatigue in laparoscopic hand manipulation [1].Hand tremor is the high frequency uncontrolled vibration of hand. One of the major causes for nonpathological hand tremors is the muscle fatigue. Effect of surgeons' hand tremor in laparoscopic surgery is crucial as it affects the accuracy of the surgery [2]. Differentiating intended motion and the high frequency tremor is important in surgical tool manipulation and positioning, also it can be used as metric for trainee surgeons. As the time progresses, the hand movements are affected by the muscle fatigue induced tremor. In this work, we have tried to address these issues by proposing a sensitive detection and analysis of muscle fatigue and induced tremor level. The broader aspect of the work will be dedicated to the compensation of the tremor effect on tool positioning.Characteristic analysis of fatigue induced tremor has been studied by several groups. Frequency domain, time domain, and time–frequency based analyses have been reported previously and all of them have highlighted the need of a good qualifier for the fatigue induced tremor [3]. EMD has a nonparametric approach for decomposition with a posteriori defined basis that is adaptive in nature [4], which makes it a legitimate choice in this scenario.In the proposed method, the surface electromyography (SEMG) and accelerometer (Acc.) data were decomposed into elementary modes using EMD with varying frequency–amplitude subsets of the signal called intrinsic mode functions (IMFs) [5]. A Fourier transformation at this stage will reflect the IMF spectral characteristics. Higher order IMF prominently captures the phenomenon of muscle fatigue and the related hand tremor activities for all the subjects. The SEMG-IMFs are used for detection of the muscle fatigue and the Acc. data IMFs were used for analysis and detection of fatigue induced tremor.EMD process starts with finding the IMFs from the raw signal. In the beginning, upper and lower envelops were found by interpolating local peaks and crests of the raw signals. A residue was calculated using the norm difference of the original signal and the envelop mean. Norm of difference calculation process was repeated until the following conditions are satisfied:Once the residue is found to obey the criteria, it is considered as the first IMF. Then, the first IMF is deducted from the raw signal and the remaining part is used to find the next IMF. This process is repeated while checking the monotonicity of the signal. The limiting value was adopted from the literature as 0.03 [6]. The final residue becomes the last component of the data subset. After the EMD, the original signal can be represented as(1)z(t)=∑i=1nxi(t)where i is the number of modes of the decomposed signal. In reality, this separation is not so straight forward. The IMFs are zero mean AM–FM (amplitude-frequency modulated) signals. This mode possesses different spectral characteristics. As the lower modes contain higher frequency part of the signal, a tremor minimization is possible by computing(2)z¯(t)=∑i=mnxi(t)with m > 1. Power spectral density (PSD) of (Z(t)-Z¯(t)) will ensure the quality of the compensation process.Hand tremor was recorded by placing inertial sensors (Acc.) on the concerned forearm and palm. Related muscle activi