Objectives: To complement current diagnostic techniques, here we present a set of objective metrics, which depend exclusively on the statistical signatures of variability inherently present in physical movements. These metrics permit assessment of the same individual over time.
Methods: In the classroom setting a computer interface was used to adapt tasks from the children’s curricula and measure natural movements. A touch screen, electromagnetic sensors (Polhemus Liberty, 240 Hz), and video recordings were synchronized and all behavioral events time-stamped and logged by the interface for later analyses. Discrete trial instruction tasks (match-to-sample) were used to evoke pointing movements familiar to the children (11 children with ASD, mean age = 9.9 years and 6 typically developing (TD) children, mean age = 4.75 years). The cognitive load of the tasks was systematically manipulated. We assessed the effects of changes in cognitive load on the hand pointing trajectories of the children. Their pointing motions were taken as a “wholesome unit” formed by a segment actively directed to the goal (the sample or target) and a segment passively carried along in transition to another active motion. The frequency distributions of the absolute maximum velocity values from the goal-directed and spontaneous movements were obtained for each child across hundreds of trials, both for novice and well-practiced trials. The former were identified with multimodal speed profiles (multiple acceleration and deceleration phases). The latter were identified with unimodal speed profiles (a single acceleration and a single deceleration phase). The continuous probability Gamma distribution was used to fit the empirical distributions using maximum likelihood estimation. Each child was represented as two data points (goal-directed and spontaneous motion) in the Gamma-distribution phase space (shape and scale).
Results: The sample clustered into different classes: The ASD children fell towards the exponential range of the Gamma-distribution while the TD children localized in the normal range of the Gamma. A linear polynomial trend best fit the data. Unfolding these clusters by performing the same fitting procedure for each individual revealed a continuum of values spanning the full Gamma range that unambiguously set apart the children with ASD (exponential-range) from the TD children (normal-range) independent of age. The scatter was characterized by a power relation. During the cognitive-dependent motor learning the points shifted to different degrees for each child. We report on this shifting and indicate its relationship with scales of current diagnostic and assessment tools used for ASD.
Conclusions: It is possible to objectively quantify cognitive-dependent motor learning in children with ASD relative to TD children longitudinally in the classroom environment. We discuss implications for the quantification over time of permanent vs. transient gains in each individual child.
See more of: Cognition and Behavior
See more of: Symptoms, Diagnosis & Phenotype