Predictions of Stroke Status from Kinematic Data
Can kinematic data alone predict if a patient has had a stroke or not?
John Krakauer and Jeff Goldsmith are studying how stroke affects motion integrity by collecting and analyzing repeated motions drawn to eight targets (above: the motions of a control (left) and a severe stroke patient (right)).
In a study aiming to understand the association of stroke and motion integrity, stroke patients as well as controls were prompted to make reaches to eight different targets on a flat, two dimensional plane. The repeated reaches to different targets in random order are recorded by a device that tracks the trajectory of the reach. The data are very structured and complex, with reaches being clustered within subject, within hand, and within target. As well, each reach is parameterized by an x-coordinate and y-coordinate over the elapsed time of the reach from origin to endpoint. Being able to predict strokes from controls using only the kinematic reaching data is a first step in being able to predict other brain-related disorders using kinematic data as a diagnostic. For instance, from kinematics alone could we predict the onset of Alzheimers and Parkinsons?