PFI-TT: Novel Artificial Intelligence Approach for Automatic Identification of Genetic and Neuroimaging Markers of Autism Spectrum Disorder

创新伙伴关系-技术转化（PFI-TT）项目的更广泛影响/商业潜力是将诊断为自闭症谱系障碍（ASD）的患者的年龄减少到大约6个月，并产生一个个体属于自闭症谱系的详细图片。这些目标将通过一个支持人工智能（AI）的软件框架来实现，该软件框架将分析有自闭症风险儿童的大脑结构和其他医疗信息。该系统可能会减轻获得自闭症诊断的负担，在目前的临床实践中，这需要花费大量的时间和数千美元。ASD症状与大脑区域的详细映射可能有助于儿科医生和其他专家更好地交流他们的诊断结果，并告知他们的治疗计划。此外，减少自闭症的诊断年龄将给父母和照顾者提供更大的时间窗口来应用早期强化行为干预，这是已知的改善自闭症儿童的结果。该项目旨在建立一个基于多模态脑成像和基因组风险因素的客观指标的自闭症诊断计算机辅助诊断（CAD）系统。儿童自闭症的诊断目前依赖于对儿童行为的主观评估。诊断过程可以早在一两岁时就开始，但它会持续到3-4岁的随访观察，直到最终的自闭症诊断。这个过程总共要花费5000到7000美元。所提出的CAD系统有望以目前成本的一小部分产生快速诊断。它试图确定大脑解剖学的特定方面（从结构磁共振成像[MRI]），以及大脑连接（从功能和扩散MRI），这些方面将与特定的ASD行为亚型相关。将这些神经学数据与患者基因组中的ASD相关变异相结合进行评估，将产生与ASD症状学有关的大脑区域和神经回路图的详细概况。这张地图将使用深度机器学习来训练系统，对一组高风险婴儿进行回顾性队列训练，这些婴儿在一岁前接受了脑成像，后来被诊断为自闭症。系统将根据一个独立的数据集进行验证。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this Partnerships for Innovation - Technology Translation (PFI-TT) project is to reduce the age of patients diagnosed with autism spectrum disorders (ASD) to approximately six months and produce a detailed picture of where an individual falls on the autism spectrum. These aims will be accomplished via an Artificial Intelligence (AI)-enabled software framework that will analyze brain structure and other medical information from children at risk for ASD. The system may reduce the burden of obtaining a diagnosis of autism, which takes a significant amount of time and thousands of dollars under current clinical practice. The detailed mapping of ASD symptoms to brain regions may help pediatricians and other specialists better communicate their diagnostic findings and inform their plans for treatment. Furthermore, reduction in age at which autism can be diagnosed will give parents and caregivers a greater window of time to apply early intensive behavioral interventions which are known to improve outcomes in autistic children.The proposed project aims to produce a computer-assisted diagnostic (CAD) system for autism diagnosis based on objective metrics derived from multimodal brain imaging and genomic risk factors. Pediatric autism diagnosis currently relies on subjective evaluations of child behavior. A diagnostic process can begin as early as one or two years of age, but it continues with follow-up observations through age 3–4 years till a final autism diagnosis is rendered. This process can cost $5000–$7000 in total. The proposed CAD system is anticipated to produce a rapid diagnosis at a fraction of current cost. It seeks to identify specific facets of brain anatomy (from structural magnetic resonance imaging [MRI]), and brain connectivity (from functional and diffusion MRI) that will correlate with specific behavioral subtypes of ASD. Evaluating these neurological data in concert with ASD-related variations in the patient’s genome, will produce a detailed profile of brain regions and neural circuit maps implicated in ASD symptomatology. This map will be accomplished using deep machine learning to train the system on a retrospective cohort of high-risk infants, who underwent brain imaging prior to one year of age and were later diagnosed with autism. The system will be validated against an independent dataset.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.