Multimodal AI-based Diagnosis of Autism Spectrum Disorder (ASD)

自闭症谱系障碍（ASD）是一种终身障碍，影响人们的感知和与世界的互动，它影响着世界各地的大量人口。这个令人兴奋的项目旨在使用人工智能和多模态数据，提供补充信息，以提高ASD诊断的准确性，这些诊断仍然依赖于传统的基于观察的方法，如自闭症诊断观察时间表，即使对于自闭症患者，也可能以评估结果的不确定性结束。此外，这些方法大规模实施成本高昂，需要训练有素的专业人员。本项目的主要目标是（1）开发一种多模态深度学习模型，可以从ASD患者的行为特征中学习，以帮助准确诊断。为此，其他目标是：（2）进行实验以收集多模态数据（例如，视觉和听觉数据、社交信号和EEG测量），以研究使用多模态数据对ASD诊断的影响，并调查不同模态中特征之间的相互关系，以及（3）调查自闭症谱系严重程度的特征这一雄心勃勃的项目超越了目前相关文献中的最先进水平。它将允许学生与多学科研究小组合作，并扩大他/她对医疗保健，人工智能，机器人和心理学所面临的挑战的理解。

Autism Spectrum Disorder (ASD) is a lifelong disorder that affects people's perception and interaction with the world, and it affects a large population worldwide. This exciting project aims to use AI and multimodal data, providing complementary information, to enhance ASD diagnosis accuracy that still depends on traditional observation-based approaches, like the Autism Diagnostic Observation Schedule, that may end with uncertainty about the assessment outcome even for those with autism. Moreover, these approaches are costly to conduct on a wide scale and require trained professionals.The main objective of this project is to (1) develop a multimodal deep learning model that can learn from features characterizing the behaviour of persons with ASD to assist in accurate diagnosis. To this end, the other objectives are; (2) conduct experiments to collect multimodal data (e.g., visual and audio data, social signals, and EEG measures) during interaction with peers, professionals, and social robots to study the effects of using multimodal data on ASD diagnosis, and to investigate the inter-relationships between features in different modalities, and (3) investigate features that characterize severity levels on the autism spectrum (from mild to severe levels).This ambitious project goes beyond the current state-of-the-art in the related literature. It will allow the student to work with a multidisciplinary research group and broaden his/her understanding of the challenges facing healthcare, AI, robotics, and psychology.