Xin Li ^1* Mindi Ruan ² Na Zhang ² Xiangxu Yu ³ Wenqi Li ² Chuanbo Hu ¹ Paula J. Webster ⁴ Lynn K. Paul ⁵ Shuo Wang ³

• ¹ University at Albany, Albany, United States
• ² West Virginia University, Morgantown, West Virginia, United States
• ³ Washington University in St. Louis, St. Louis, Missouri, United States
• ⁴ Congressionally Directed Medical Research Programs, Fort Detrick, MD 21702, United States
• ⁵ California Institute of Technology, Pasadena, California, United States

Early and accurate diagnosis of autism spectrum disorder (ASD) is crucial for effective intervention, yet it remains a significant challenge due to its complexity and variability. This study introduces a novel machine-learning (ML) framework that advances ASD diagnostics by focusing on facial micro-expressions. Micro-expressions are rapid, involuntary facial movements indicative of underlying emotional states. We applied cutting-edge algorithms to detect and analyze these micro-expressions from video data, aiming to identify distinctive patterns that could differentiate individuals with ASD from typically developing peers. Our computational approach included three key components: 1) micro-expression spotting using Shallow Optical Flow Three-stream CNN (SOFTNet), 2) feature extraction via Micron-BERT, and 3) classification with majority voting of three competing models (MLP, SVM, and ResNet). Despite the sophisticated methodology, our experimental results suggest that the ML framework's ability to identify ASD-specific patterns reliably was limited by the quality of video data. This limitation raised concerns about the efficacy of using micro-expressions for ASD diagnostics and pointed to the necessity for enhanced video data quality. Consequently, our research has provided a cautious evaluation of micro-expression diagnostic value, underscoring the need for further advancements in behavioral imaging and multimodal AI technology to leverage the full capabilities of ML in an ASD-specific clinical context.