Highly scalable technology-assisted differential diagnostics of ASD

Diagnosing autism spectrum disorder (ASD) in adulthood is time-consuming and markedly complicated by the requirement to distinguish between ASD and differential diagnoses also associated with social interaction difficulties, such as Borderline Personality Disorder (BPD) - a distinction for which currently no valid screening or diagnostic tool exists. While technology-assisted diagnostics (TAD) has emerged, existing algorithms have focused on classifying between ASD and no diagnosis, not fully addressing clinical reality. Therefore, we assessed the feasibility of TAD for differential diagnostics by classifying between ASD and BPD. We extracted features from live reciprocal conversations, allowing us to capture the core area of defining symptoms for both conditions: social interactions. We collected a rich, multimodal dataset of dyads using hyperrecording to capture different communication channels in a time-locked manner (speech, facial expressions, motion). Then, we trained support vector machines to classify between dyad types (ASD-involved, BPD-involved and comparison dyad). Stacking several models containing conceptually related features, our algorithm achieves a near 82% of balanced accuracy, solely based on 20 minutes of conversation. These results show the immense potential of TAD for differential diagnostics: data collection only requires microphones and webcams while feature-extraction is automated, making this approach highly objective, scalable and user-friendly.