Abstract
In clinical application of FES, the method, in which the standard stimulation data are made based on the EMG of normal subjects and adjusted to each patient, has been used. However, this method was burdensome to normal subjects, patients, and medical staff. The final goal of this study is to develop a practical musculoskeletal model which is easily applied in clinic and to automatically generate the stimulation data for a specific patient with this model. The purpose of this paper is, before the final goal, to show that the stimulation data generation method using the model simplified for clinical application is useful as an alternative of the EMG based method. We set the paraplegics' standing-up motion unassisted by the upper extremities as the first example problem, because standing-up motion is thought to be the starting point of motor function restoration of paraplegics. Standing-up motion pertains to the movement of each segment in the sagittal plane, so that 3-link model in the sagittal plane was selected as a skeletal model. Previous musculotendon model for FES of other groups often ignored the difficulty of identifying the model parameters, which made the model impractical in clinical application. Therefore, we made the musculotendon model easier to be clinically applied by simplifying it to be described only with the parameters available from noninvasive measurements. Then we generated the stimulation data for standing up without arm-assistance by dynamic optimization method using the model simulation. The resulted stimulation data and estimated motion were compared with the EMG patterns and motions of normal subjects' standing up, respectively. The joint torque trajectories developed by model simulation were similar to those of a normal subject calculated by Bajd, et. al, the generated stimulation data were roughly in agreement with the normal subjects' EMG, and the cost function is normally minimized in dynamic optimization process and properly incorporated in the stimulation data. From these, we may conclude that the developed model is proper in principle, and that the method of this paper is useful as one for generating standard stimulation data. The developed system is expected to become more practical through refinement of the cost function and constraints of optimization and through improvement and extension of the model considering clinical applicability.
