Automated Detection and Characterization of Stereotypical Motor Movements and Their Relation to Cardiovascular Arousal

Background: One of the core diagnostic features of autism spectrum disorder (ASD) is engagement in stereotypical motor movements (SMM), although the etiology of this repetitive behavior is unknown. Since the 1960s, it has been hypothesized that SMM serve a homeostatic regulation function, and thereby a coupling mechanism to cardiovascular arousal. However, to date, surprisingly few reports explicitly assess cardio-somatic coupling and SMM in ASD.

Objectives: Examine naturally emitted SMM concomitantly with peripheral nervous system measures for evidence of cardio-somatic coupling.

Methods: SMM recorded via video, three-axis accelerometry, and simultaneous electrocardiographic signals were obtained in a convenience sample (n=10) of children and young adults with moderate to profound ASD to assess cardiovascular responses at or around the onset of naturalistically occurring stereotypy. RR intervals were extracted via Vivometrics software, visually inspected, and hand-corrected using CardioEdit software. Corrected RR intervals were then time-aligned with video annotations, resampled to 5Hz, and subsequently analyzed both as short sections defined by time (designated window analysis) or by SMM onsets (designated point analysis).

Results: Twenty-four sessions varying in length were analyzed (n=24, mean=25.75 min, SD=8.22, range [8.18-49.81]). Most participants were recorded for approximately 30 to 90 minutes each over multiple sessions. Agreement between annotators in the video codes ranged from k = 0.72–0.92 (M = 0.84), suggesting high inter-rater reliability for observed SMM type, start, and end times. Rocking and flapping were the most commonly observed SMM (n=22 and n=20 sessions, respectively). Without any restrictions on activity or duration, windows containing any SMM were compared to windows without SMM. Averaged across all participants, there was very little change in RR interval (mean ∆=-6.8ms, SD=20.2ms) or RSA (mean ∆=0.03, SD=0.16 ln(ms²/Hz)). We then extended our analyses to evaluate physiological response differences in type and duration of SMM at the per participant session level. Differences in heart interval and RSA were computed between rocking vs. no SMM, flapping vs. no SMM, and mixed stereotypy vs. no SMM. Additionally, responses were separated by different levels of SMM engagement (any, moderate, heavy). Results indicated a moderate effect on RSA with rocking and a large effect on heart interval with flapping. Finally, instantaneous heart rate changes for one, two, and four second onset latencies produced nearly identical cardiac signatures (5^th order polynomial), including the same pattern of comparatively rapid acceleration followed by a slower deceleration observed at all SMM onsets and overall SMM types.

Conclusions: Cardiovascular analyses reveal that both stereotypical body rocking and hand flapping produce a strikingly similar pattern of acceleration and deceleration unrelated to physical demands associated with the movements themselves. However, despite this consistent cardiovascular signature observed across individuals, neither type of SMM provoked directional change in heart rate variability. The data, methods, and results from the current study can be applied and extended in future work to more systematically evaluate the potential role SMM plays in homeostatic regulation, providing potentially important insights for understanding and intervening upon a core symptom of ASD.