Scholarlink

<p>Socio-motor behaviours, such as interpersonal motor synchrony (IMS), are widely thought to be atypical in autism but remain difficult to quantify objectively across the full spectrum of functioning and development. Existing studies have largely relied on controlled paradigms with seated, older participants, constraining movement and limiting insight into naturalistic interactions, particularly in young children. We analysed 69 autistic and 32 typically developing (TD) children aged 14–46 months during free-play segments of ADOS-2 diagnostic sessions. To measure IMS between child–assessor dyads, we developed a two-stage pipeline integrating identity tracking (SAM 2.1) and pose estimation (RTMPose) to extract frame-level skeletal coordinates. IMS was operationalised using dynamic time warping (DTW). Dyads involving autistic children exhibited lower IMS than TD dyads, in both trunk and head regions. Members of dyads with autistic children also showed greater activity and speed variability. Critically, IMS reductions were associated with movement characteristics of both partners in TD dyads, but primarily with child movement in dyads with autistic children, suggesting a more asymmetric coordination pattern. Lower IMS was further associated with higher autistic traits and lower cognitive functioning. These findings highlight the potential of pose estimation and DTW as scalable, non-invasive tools for quantifying atypical IMS and supporting more precise autism phenotyping in naturalistic settings.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Pose estimation analysis of free-play reveals atypical interpersonal motor synchrony in autism

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月14日

查看全文
  • Sara Akhavan,
  • Nada Kojovic,
  • Thibaut Chataing,
  • Julie Segers,
  • Thomas Maillart,
  • Marie Schaer

摘要

Socio-motor behaviours, such as interpersonal motor synchrony (IMS), are widely thought to be atypical in autism but remain difficult to quantify objectively across the full spectrum of functioning and development. Existing studies have largely relied on controlled paradigms with seated, older participants, constraining movement and limiting insight into naturalistic interactions, particularly in young children. We analysed 69 autistic and 32 typically developing (TD) children aged 14–46 months during free-play segments of ADOS-2 diagnostic sessions. To measure IMS between child–assessor dyads, we developed a two-stage pipeline integrating identity tracking (SAM 2.1) and pose estimation (RTMPose) to extract frame-level skeletal coordinates. IMS was operationalised using dynamic time warping (DTW). Dyads involving autistic children exhibited lower IMS than TD dyads, in both trunk and head regions. Members of dyads with autistic children also showed greater activity and speed variability. Critically, IMS reductions were associated with movement characteristics of both partners in TD dyads, but primarily with child movement in dyads with autistic children, suggesting a more asymmetric coordination pattern. Lower IMS was further associated with higher autistic traits and lower cognitive functioning. These findings highlight the potential of pose estimation and DTW as scalable, non-invasive tools for quantifying atypical IMS and supporting more precise autism phenotyping in naturalistic settings.