Patient-prosthesis interaction: : Control through the healthy leg

Abstract

The loss of a lower limb is an irreversible and traumatic event in the life of the affected person. Although modern prostheses can restore the function of the missing body parts to a high extent, commercial state-of-the-art lower limb devices neither measure nor incorporate environmental information for intent recognition or for device control. This inevitably leads to the problem that the patient adapts to the behavior of the prosthesis rather than the system to the amputees’ needs. Unphysiological gait patterns and increased compensatory movements are just some of the known consequences. This thesis offers a concept for improving patient-prosthesis interaction. The device control is enhanced by measuring and evaluating the position of the unimpaired residual contralateral leg. This leads to an even more lifelike replication of the physiological gait pattern. First, an extensive literature survey was conducted to find out which modalities of environmental sensors are already being used and how they improve the control of lower limb prostheses. During this review, five control approaches were identified, as to how “next generation prostheses” could be optimized. Overall, there is a clear trend towards more upcoming terrain or object estimation, with the basic idea of delivering switching probabilities between different activities. Most relevant to this work, however, was the conclusion that even a single sound-leg measurement can significantly reduce the error rate in detecting the amputees’ intent correctly. Based on these findings, a first depth camera-based system was developed. This contralateral limb tracking (CoLiTrack) approach, combines a single depth camera with an inertial measurement unit, both mounted on the ipsilateral leg, which is able to estimate the shank axis of the contralateral side. Initially, the scene captured by the camera was transformed into a stabilized world coordinate system. In order to achieve real-time performance, the subsequent shank-estimation process was split into two less computationally intensive steps: First, circular models were fitted against 2D projections of the input using the iterative closest point algorithm. Second, the final shank axis was determined by applying the random sample consensus method. In three experiments, data from five able-bodied subjects was evaluated. The results demonstrated a trackability of the shank axis over one sixth of the entire human gait cycle for dynamic situations. In order to overcome the limitations of the previous depth camera-based approach, a second contralateral limb ranging (CoLiRang) system was developed. Using four novel ultrasonic time of-flight sensors on the ipsilateral leg, the position of the other contralateral leg was estimated. Initially, each sensor measured the respective distances to the contralateral side. These distances were then triangulated to determine the directional information of the other leg. In order to evaluate the system as a whole, several tests were performed and experiments with two healthy participants were conducted. The results showed a mean triangulation deviation of less than 30mm and a divergence in detecting the moment of passing even below 1°. Most importantly, this novel approach was able to track the state of the other leg correctly in dynamic situations throughout the entire human gait cycle.Finally, the ultrasonic-based concept was integrated into an enhanced seeing prosthesis (SEP) system. For the first time, it was possible to control the damping behavior, walking resistance of the device, via the state of the patient’s unimpaired contralateral residual leg in real time. In order to evaluate the novel system, a prospective pilot clinical study was designed, approved, and conducted with five transfemoral amputees. Closed-loop optimization sessions were conducted first, followed by a clinical biomechanical gait analysis with each participant. The results revealed a more physiological gait pattern and a distinct facilitation of the remaining musculoskeletal system for yielding activities. In particular, the interception on the contralateral healthy leg was reduced on average by about 25% for going down the ramp and even by about 40% for the staircase task, respectively.This work has demonstrated that environmental sensing technologies can successfully improve the patient-prosthesis interaction. Taking these findings into account for further development, prostheses of the next generation would be able to truly adapt to patients’ needs.