Development of a Method for Creating Stimuli Using Deepfake Videos and a Video Self-Modeling Study Using the Generated Videos in Children with Autism Spectrum Disorder
ディープフェイク映像を使った自閉スペクトラム症児へのビデオ・セルフ・モデリングの研究
A study of video self-modeling using deepfake videos for children with autism spectrum disorder
[Common requirements for typically developing children and children with autism spectrum disorder] Written proxy consent for the child's participation; no facial injury or skin condition that interferes with filming. [Typically developing children] Aged 5 years or older and younger than 7 years at consent; no history of a neurodevelopmental disorder reported by the proxy; able to participate in the procedures based on proxy report and researcher observation during explanation and filming practice. [Children with autism spectrum disorder] Aged 5 years or older and younger than 7 years at consent; diagnosed with autism spectrum disorder; no intellectual disability; able to perform motor imitation, object-use imitation, instruction following, and video viewing; able to attend to the video for at least 3 minutes; written proxy consent for the child's participation and the proxy questionnaire. [Proxy group] Proxy for a participating child with autism spectrum disorder; written consent and ability to complete the questionnaire.
[Typically developing children] Judged unsuitable for participation by an investigator. [Children with autism spectrum disorder] Receiving training at a rehabilitation or support facility for a behavior that is the same as or similar to the target behavior and may affect evaluation of the intervention; or judged unsuitable for participation by an investigator. [Proxy group] Judged unsuitable for participation by an investigator.
For children with autism spectrum disorder, a multiple-baseline design across participants will be used. After a 3- to 7-week baseline, video self-modeling using a deepfake video in which the child appears to perform the target behavior will be delivered once weekly for 8 to 12 sessions, followed by a 4-week follow-up. Each session will last about 1 hour, and the target behavior will be individually selected based on screening results and daily-life needs. Typically developing children will participate in one approximately 30-minute filming session. Proxies will complete a post-intervention questionnaire.
介入コード / Code
D001521
介入キーワード /Keyword
ビデオ・セルフ・モデリング,行動療法,ディープフェイク映像
video self-modeling,behavior therapy,deepfake video
1) Typically developing children: cosine similarity of facial expressions between original and deepfake videos based on 17 action units extracted with OpenFace 2.2.0, assessed after filming and video generation. 2) Children with autism spectrum disorder: target behavior occurrence rate, assessed by partial-interval recording from session videos during a 3- to 7-week baseline, 8 to 12 weeks of weekly intervention, and a 4-week follow-up.
副次的な評価項目 / Secondary Outcome(s)
1) 自閉スペクトラム症児群におけるタブレット領域及び人物領域への総注視時間と注視割合。介入セッション中にTobii Pro Glasses 2で計測する。2) 介入後に代諾者アンケートで評価する、標的行動の変化及び日常生活での有用性。
1) Total gaze duration and proportion of gaze directed to the tablet and person areas in children with autism spectrum disorder, measured during intervention sessions using Tobii Pro Glasses 2. 2) Changes in the target behavior and usefulness in daily life, assessed by a post-intervention proxy questionnaire.