Characterization of the Statistical Signatures of Micro-Movements Underlying Natural Gait Patterns in Children with Phelan McDermid Syndrome: Towards Precision-Phenotyping of Behavior in ASD

Abstract

Background: There is a critical need for precision phenotyping across neurodevelopmental disorders, especially in individuals who receive a clinical diagnosis of autism spectrum disorder (ASD). Phelan-McDermid deletion syndrome (PMS) is one such example, as it has a high penetrance of ASD. At present, no biometric characterization of the behavioral phenotype within PMS exists.Methods: We introduce a data-type and statistical framework that permits the personalized profiling of naturalistic behaviors. Walking patterns were assessed in 30 participants (16 PMS, 3 idiopathic-ASD and 11 age- and sex-matched controls). Each individual's micro-movement signatures were recorded at 240 Hz. We empirically estimated the parameters of the continuous Gamma family of probability distributions and calculated their ranges. These estimated stochastic signatures were then mapped on the Gamma plane to obtain several statistical indexes for each child. To help visualize complex patterns across the cohort, we introduce new tools that enable the assessment of connectivity and modularity indexes across the peripheral network of rotational joints.Results: Typical walking signatures are absent in all children with PMS as well as in the children with idiopathic-ASD (iASD). Underlying these patterns are atypical leg rotational acceleration signatures that render participants with PMS unstable with rotations that are much faster than controls. The median values of the estimated Gamma parameters serve as a cutoff to automatically separate children with PMS 5–7 years old from adolescents with PMS 12–16 years old, the former displaying more randomness and larger noise. The fluctuations in the arm's motions during the walking also have atypical statistics that separate males from females in PMS and show higher rates of noise accumulation in idiopathic ASD (iASD) children. Despite high heterogeneity, all iASD children have excess noise, a narrow range of probability-distribution shapes across the body joints and a distinct joint network connectivity pattern. Both PMS and iASD have systemic issues with noise in micro-motions across the body with specific signatures for each child that, as a cohort, selectively deviates from controls.Conclusions: We provide a new methodology for precision behavioral phenotyping with the potential to use micro-movement output noise as a natural classifier of neurodevelopmental disorders of known etiology. This approach may help us better understand idiopathic neurodevelopmental disorders and personalize the assessments of natural movements in these populations.

Motor noise is rich signal in autism research and pharmacological treatments

“…To overcome these potential issues we median-ranked the scores and used these automatic groupings to pool their underlying micro-movements waveforms5. Individuals with ASD were thus grouped by their ADOS scores, ranked above and below the median values and compared to CT. Social affect (SA), repetitive and restrictive behaviours (RRB) scores, and Severity Scores (SS) were based on research-reliable administration of ADOS (scores reported were computed based on revised Gotham algorithms2526; see Methods for details).…”

“…1C). Excess or deficits in involuntary motor variations relative to normative scales is undesirable, and has been found in the context of goal directed reaches3, decision making4 and gait patterns5 across various clinical populations with pathologies of the nervous system, including ASD3678910. Importantly, subtle fluctuations in the movement signal generate and carry new signals in a returning afferent stream: a form of re-entrant sensory feedback from the PNS to the CNS11, putatively conveying sensory feedback linked to self-generated movements.…”

“…To model them, we build on previous research6 whereby the amplitudes and inter-spike interval times are modeled as independent and identically distributed (iid) random variables following a Gamma distribution. This framework has proven amenable to computational tractability, facilitating both inference462324 and interpretation of the results. As such, the Gamma process is used here in combination with a waveform (coined “micro-movements”) representing the fluctuations in amplitude and timing of the spike trains derived from the rate of change in head positions and orientations.…”

Neonatal Diagnostics: Toward Dynamic Growth Charts of Neuromotor Control

“…To be more precise, we treat the spikes in acceleration as spike trains of random amplitudes and random times. To model them, we build on previous research whereby the amplitudes and inter-spike interval times are modeled as independent and identically distributed (iid) random variables following a Gamma distribution (18, 19, 23, 24). As such, the spike trains are the input to a Gamma process and the empirically estimated Gamma parameters are dynamically tracked on various parameter spaces to uncover self-emerging patterns.…”

“…The various degrees of variability associated with the various layers of neuromotor control in Figure 2A have different noise-to-signal ratios (17–19). This type of information, echoed back to the brain through reafferent loops (20), may play a critical role in early stages of neurodevelopment.…”