User:Bokkyu Kim/sandbox

Assignment 4

Figure 1. The trend of increase in publications in 'Motor Control' area related to stroke

The number of publication in Motor Learning after Stroke area has been increased recently. I think there are few reasons why this area is growing.

First, there were few key research papers which open the door for this area. Nudo et al. (1996) suggested that functional reorganization occurs in the motor cortex of adult monkeys after a focal ischemic brain injury after physical training. After this paper was published, many studies in neurologic rehabilitation domain focused on the neuroplasticity after stroke. Also, Krakauer (2006) published a nice review paper about the relationship between motor learning and functional recovery after stroke. Before these papers were published, most of rehabilitation treatment for individuals after stroke was related to reducing spasticity and improving range of motion. On the other hand, the concept of rehabilitation treatment for post-stroke individuals changed to more active training to facilitate neuroplasticity and skill acquisition.

Second, emerging new technology in neuroimage could be a possible reason why the motor learning after stroke area is growing. Neuroimaging techniques such as function Magnetic Resonance Imaging (fMRI) and Transcranial Magnetic Stimulation (TMS) allow to explore the brain, and help to research the changes in the brain after rehabilitative training for individuals after stroke. These neuroimaging techniques have been used since early 1990's, but researchers started using these techniques more for motor learning study in early 2000's.

Assignment 3

The first research project which I have been involving is NEXUS project supported by Rehabilitation Engineering Research Center (RERC) at University of Southern California. The purpose of this project was to understand the differences in visuomotor control of older adults between real and virtual environments. Before I joined the project, other researchers already collected the goal-directed arm reaching kinematic data and other types of data, such as surface electromyography of lower extremity muscles and response time data, in real and virtual settings from thirty older individuals. The followings are brief explanation of the experimental procedure: Participants were asked to perform goal-directed arm reaching movements to real (tennis balls on the end of sticks) and virtual (Jewels in the Jewel Mind Video Game) targets. In addition, there were two postural demand conditions (reaching while standing stationary, and reaching while stepping). In virtual target condition, the targets appeared at one of eight locations, equally distributed around a ring parallel to the participant’s frontal plane. Within a ‘block,’ each of the 8 targets was presented. Participants had seven blocks of reaches in each condition for a total of 56 reaches per condition, with the pseudorandom target order held constant. In real target condition, an experimenter placed the tennis ball at one of the eight equally spaced target locations at a set distance in front of each participant. The Microsoft Kinect camera and accelerometers were used to record reaching kinematics during the tasks.

My research plan for the project was to find some meaningful results from the plenty of data to understand the difference in motor control of older adults between real and virtual environments among the collected various data sets. First I started analyzing the kinematic data recorded with the Kinect camera. The Kinect camera captured the distance between the camera and the subject. The Kinect camera recorded reaching kinematics at a sampling rate of 11–14Hz. The x-,y-, and z-coordinates of each wrist were captured using a virtual marker on each wrist joint. The raw Kinect data were converted to the C3D file format and low-pass filtered using a Butterworth filter with a cutoff frequency of 3Hz. Finally, the filtered data were resampled at 12Hz (the mean sampling frequency of the Kinect during data acquisition). After filtering and resampling, the tangential velocity of the virtual wrist marker was computed using the processed position data. After the data processing, I develop some codes for implementing automatic routines using Visual 3D Software to mark movement onset, peak velocity, and movement offset on each trial. Also, to define movement onset and offset, I had to review a number of literatures related to human arm reaching kinematic analysis. This is because the kinematic data was captured with low sampling rate, the general method to define movement onset and offset was not appropriate for my study. So I searched some studies which used low sampling rate to record kinematic data, and I adopted a definition of movement onset and offset from Messier & Kalaska (1999).

For each reach, kinematic measures associated with the planning and control of reaching movements (peak velocity, PV, time to peak velocity, TTPV, and movement time, MT) of the wrist trajectory were computed. Usually, peak velocity represents pre-planning, and time to peak velocity represents on-line fast feedback of movements. In this study, participants scaled PV with distance in both virtual and real settings, indicating similar pre-planning of goal-directed reaching movements. When compared to real targets, participants demonstrated increased scaling of TTPV when reaching to virtual targets, indicating more use of online fast feedback in virtual settings while standing, but not when stepping. These results may indicate that the visuomotor challenge of virtual environments does not alter pre-movement planning, but does interact with the use of on-line fast feedback for movement correction. When I looked at each participant’s individual data, some participant showed different shape in velocity profile between real and virtual settings. For example, the slope of velocity graph in real settings is steeper than that in virtual settings at the initiation and the end of reaching movements. I think this is because of the uncertainty of the target location in virtual settings, and this uncertainty in virtual environments can reduce the acceleration at the initiation and deceleration at the end of reaching movements. So I am analyzing the acceleration profile of each reaching movement. Due to the low sampling rate of the Kinect camera, I was unable to differentiate twice the position data from the Kinect to calculate the acceleration. So I am trying to use the acceleration data from accelerometers, and developing matlab codes to analyze acceleration data.