Introduction: Myoelectric prosthetic limbs have been well accepted by upper-limb amputees for many years, and advances in myoelectric control systems have increased the popularity of these devices. Pattern recognition–based control of powered upper-limb myoelectric prostheses offers a means of extracting more information from the available muscles than do conventional methods and therefore can be used to increase the number of functions in an artificial limb. Traditionally, surface electromyography (EMG) has been used to investigate muscle activation patterns to determine which areas of a residual limb would be appropriate for electrode placement and control. High-density EMG (HDEMG) systems have allowed for noninvasive collection of myoelectric signals (MES) from many closely spaced electrodes. The data obtained can be examined through the use of “color maps,” which provide a visual indication of the distribution and intensity of muscle activation. This technology is particularly suited for those with limited muscle physiology due to injury or loss.
Materials and Methods: To further understand the activation patterns of amputees, this work focused on examining two types of contractions to determine the relationship, if any, between the color maps produced from HDEMG data and classification accuracy used for pattern recognition control. Understanding this relationship may help to develop better clinical training protocols for prosthesis users and also identify those individuals who would be the most suitable candidates for myoelectric prostheses. An HDEMG system (REFA, TMS International) was used to evaluate two common hand movements (“hand open” and “hand closed”) at a self-selected medium contraction level. Four transradial amputees (two with traumatic [TR] amputations and two with congenital [CG] amputations) participated in this study. Up to 32 surface electrodes were placed in a grid formation over the forearm region to collect data from the residual limb. The areas on the forearm that experienced muscle activity during given movements were illustrated in topographical (color) maps for each trial. The color maps were visually inspected to determine any changes in intensity (amplitude) and pattern repeatability between trials. Pattern classification accuracies were computed for both movements and compared with their corresponding color maps to observe any trends.
Results: Both color map pattern and intensity changes were noted in relation to classification accuracy; however, a quantitative relationship between the two was not determined. Although the sample size is limited (n = 4), these observations were similar for those with CG and TR amputations. The results suggest that classification accuracy differs according to both pattern and intensity changes; however, the exact relationship remains elusive.
Conclusions: Understanding this connection may help to determine which candidates are better suited for prostheses using pattern recognition versus those that may remain successful with traditional systems.