Research Proposal

Acquiring Expert-Level Skill with Telepresence Robotic Limb Via Virtual Glove: an fMRI and Learning Process Study

Keywords: fMRI, motor cortex, tool-use, skill acquisition, prosthetics, arm, action mapping, cyberpsychology

Our primary focus in the lab is on tool use with a special interest in prosthetics use, particularly in regards to adaptation to and function-exploitation of prosthetic hands and arms. To this end, we plan to study the acquisition of skill in the direction of a mechanical arm as a parallel to learning to use a prosthetic limb after amputation. With previous research supporting a goal-oriented rather than tool-specific process in the motor cortex (Tunik, Frey, and Grafton, 2005; Johnson-Frey, McCarty, and Keen, 2004), further study is necessary to determine the specifics of the process, an important question in relation to prosthetics use thanks to the apparent integration of tool into action schema, something also of more general interest regarding tool design and training overall. This obvious potential application in increasing the level of function available to those reliant on prosthetic limbs and multi-function tools lends a good degree of value to the study, hopefully aiding our ability to use effectively the increasingly complex technology we as a culture are provided. When performed, this will be my introductory research into the specific topic of interest I have pursued through college. My interest in tool use and prosthetics has been prevalent during most of my life, fueled by the frustration with limits throughout a childhood with limited health and a general fascination with the positive potential on technology that is primarily limited by individuals’ ability to use what tools they can access.
Though much of the proposed research is investigatory, seeking to provide a framework for the acquisition of tool-use and comparisons in activations in the motor cortex, the results should also provide information on the mapping of motor activity. We hypothesize that certain actions are more easily acquired in tool learning when the final action and the initiating action are most similar, such as use of a power grasp to initiate a power grasp with tongs, even if the actual actions are reversals (i.e. grasping to realize an object from reversed chemist tongs) as suggested by theories of embodied cognition (Simmering, Spencer, 2007). Similarly, we hypothesize that errors will be most prevalent when actions are most dissimilar. Furthermore, we expect our fMRI results to mirror earlier research suggesting goal-specific rather than motor-action specific activation in the motor cortex (Johnson-Frey et al, 2004). Combined, this will hopefully provide a basis from which to work in designing training programs for prosthetic-limb and other complex tool use as well as more general information regarding motor activity which may be applicable to reacquiring motor skills for stroke victims and the perception of tool function.
Being a relatively new area of interest, little prior research has been performed. Much of what has been done has focused on attention in regards to control panels or computer monitors or the use of simple tools. Thanks to this, novel research is easily performed without much risk of accidental reproduction of previous experimentation. The studies below reflect research design constructed by myself with design-refinement guidance by my supervising professor.
For the initial research, participants will be trained to reach and grasp objects using the robot via the virtual glove with a semi-arbitrary mapping configuration due to the difficulty of using the pinky to control certain actions. The degree of skill acquisition will be measured both in number and type of mistakes and by the achievement of various performance criterions, such as picking up an object or mimicking an arm posture in an image. After a final performance criterion consisting of use and coordination of each of the robot’s joints is easily achieved, subjects will be moved to an fMRI task in which the difficulty of two conditions will vary parametrically: 1) a reach condition wherein they must keep the robot’s grasping “fingers” together and make context with a target whose width is modulated geometrically, 2) a grasp condition they must reach to grasp the same targets. The geometric variations in target size allow us to modulate the difficulty of reach and grasp. They are cued aurally on a trial by trial basis whether the task is reach or grasp. The participants will begin their motions at a signal from the researcher and then have a limited period of time to achieve the goal of the condition with rest periods between attempts to achieve the goal.
This data will be analyzed in several ways: A) evaluation of the effects of the parametric modulation of target size on reach and grasp conditions separately relative to rest under the premise that the target width manipulation will selectively increase demands on control of reach or reach plus grasp depending on the task, and B) a subtractive contrast of grasp vs. reach to remove the common reach component.
The second study will be methodically similar, varying only in that instead of participants using the same mapping of finger motion to robot movement, the mapping will be randomized. Since the glove is capable of mapping any finger-movement input to any of the movements the robot arm is capable of, each participant will have a randomly assigned mapping configuration to learn. By doing so, we can control for and possibly cancel-out potential effects and quirks of specific mappings when doing more general group analysis as well as providing data for correlation analysis. In the second study, the data will be analyzed in the same way with the addition of looking at correlations between mapping configuration, robot action and component type involved in task, and error types.
If effects are found in either study one or two, a third study with fMRI scans throughout the learning process will be performed to view the process of mapping complex tool functions in the motor-cortex and to determine the relation between the use of complex tools and normal motor functions when the end goal (such as grasping) is the same. Similar data analysis to that performed in the previous studies will be performed with an added temporal component for changes and activation of learning processes over time.

Johnson-Frey, S.H., McCarty, M., & Keen, R. (2004). Reaching beyond spatial perception: effects of intended future actions on visually-guided prehension. Visual Cognition, 11, 371-399.

Simmering, VR., Spencer, JP. (2007). Carving up space at imaginary joints: Can people mentally impose arbitrary spatial category boundaries? Journal Of Experimental Psychology-Human Perception and Performance 33 (4): 871-894

Tunik E., Frey, S.H., & Grafton ST. (2005). Virtual lesions of the anterior intraparietal area disrupt goal-dependent on-line adjustments of grasp. Nature Neuroscience, 8, 505-511.