UC Irvine

An estimated one in four people will experience a stroke in their lifetime. One of the most debilitating and common consequences of a stroke is loss of sensorimotor function on one side of the body. In this dissertation we pose the question: what should we target as we develop robotic and sensor-based tools to increase paretic upper extremity use after stroke? We approach this question by identifying three gaps. First, we lack understanding of how impairment reduction can lead to use increase. Second, despite the prevalence of proprioceptive deficits after stroke and the potential role of proprioception in motor learning, there are no methods for intensely and engagingly training hand propriomotor capacity. Third, there is an unmet need for compact rehabilitation robotic devices suitable for home use. To address these gaps, this dissertation presents advances in precision medicine, rehabilitation gaming paradigms, and rehabilitation robot design. We identify responders to technology-based training by developing a model to explain changes in daily arm use after therapy, analyzing data from seven robotic clinical trials conducted by our laboratory. The identified model demonstrated that individuals with low baseline use relative to their baseline score on a common clinical measure of hand dexterity (a mismatch that we call “untapped use potential”) had a high probability of increasing use, independent of the type of study intervention.

But what of the non-responders? The model predicts that an increase in dexterity would help. We considered this finding in light of a previous study that found that finger proprioception acuity predicted participants’ ability to change their dexterity after intense robotic movement training. The problem is that few paradigms exist for retraining finger proprioception acuity. Thus, we developed a novel proprioception-targeting gaming paradigm, Propriopixels, for simultaneously training finger motor function and proprioception. Instead of displaying all game elements on screen as in a traditional video game, in the Propriopixels paradigm one of the game dimensions is conveyed to the finger with a robotic device. That is, we create a “Propriopixel” by moving the finger instead of a light pixel on screen.

We then asked, “What is the minimal robot needed to implement the Propriopixel paradigm?” Compared to more common robotic therapy designs that utilize high-cost actuators and sensors to render a continuum of impedances, we propose the design concept of a binary impedance robot that only renders the two limits – high and low impedance. It is either stiff to passively drive the Propriopixel finger, or transparent to sense active finger movements for a game input. Given the savings that solely stiff, actuated and transparent, unactuated mechanisms afford, a Propriopixels game can be cost-effectively realized with a relatively simple, binary impedance robotic device, that we demonstrate with a device called PINKIE.

We implemented Propriopixels with Proprioceptive-Pong, a game based on the classic Atari arcade game. We used the PINKIE device to study a purely passive finger movement version of Proprioceptive-Pong and the FINGER robotic exoskeleton to train an active movement version, both with unimpaired participants. We found that training with the passive version of the game yielded gains in passive proprioception acuity, while training with the active version of the game yielded gains in active but not passive proprioception acuity, suggesting a specificity of proprioceptive training principle and/or important differences between the passive proprioceptive acuity assessments deployed on each robotic device.

Following this, we studied the extent to which stroke survivors could understand and play Proprioceptive-Pong. We evaluated two methods of controlling success rates during Proprioceptive-Pong play, by either adjusting virtual game parameters only or by physically assisting players to complete successful movements. We found that a progression of game modes that gradually grew in complexity was effective for teaching Proprioceptive-Pong, and that the success control algorithm was capable of regulating success with both methods of assisting participants - virtually and physically. These results indicate that stroke survivors can understand and play Proprioceptive-Pong.