Identification of Brain Circuit Markers for Psychosis in Alzheimer's Disease by Leveraging Big Data and Machine Learning

利用大数据和机器学习识别阿尔茨海默病精神病的脑回路标志物

• 批准号: 10192576
• 负责人: Kaustubh Satyendra Supekar
• 金额: $ 23.61万
• 依托单位: STANFORD UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-09-30 至 2023-05-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10192576
• 关键词: Affect; Alzheimer' s Disease; Alzheimer' s disease patient; Artificial Intelligence; Big Data; Biological Markers; Brain; Brain imaging; Caregivers; Classification; Clinical; Clinical Psychology; Cognitive; Data; Data Set; Development; Diagnosis; Disease; Etiology; Goals; Human; Individual; Investigation; Knowledge; Lead; Learning; Machine Learning; Measures; Modeling; Neurobiology; Patients; Persons; Phenotype; Psychiatry; Psychological Transfer; Psychoses; Public Health; Reproducibility of Results; Research; Research Personnel; Sampling; Schizophrenia; Testing; Universities; Validation; Work; big-data science; clinically relevant; cognitive control; cognitive function; computer framework; connectome; deep learning; effective therapy; improved; innovation; learning strategy; multidisciplinary; neuroimaging; neuropsychiatric symptom; novel; open source; personalized diagnostics; phenotypic data; precision medicine; predictive marker; psychotic symptoms; relating to nervous system; translational neuroscience; treatment strategy

Psychotic symptoms are among the most common and persistent neuropsychiatric symptoms in Alzheimer’s disease, affecting over 50% of patients with Alzheimer’s disease. Yet, the etiology of psychosis in Alzheimer’s disease is still poorly understood and systematic investigations have been hampered by small samples and lack of reproducible findings. Critically, robust biomarkers are needed to understand the origins/progression of psychosis in Alzheimer’s disease, and to identify targets for treatment. Newly available human neurobiological data offer an unprecedented opportunity for developing robust and predictive biomarkers for psychosis in Alzheimer’s disease. The overarching goal of our proposal is to identify robust and predictive biomarkers for psychosis in Alzheimer’s disease using a novel data-driven computational framework. Specifically, we will use a transformative “Big Data” science approach combining exciting recent advances in deep learning and our recent work on quantitative dynamic brain circuit analyses with a wealth of newly available large-scale open- source brain imaging and phenotypic data from multiple consortia, as well as data we have acquired at Stanford University. To achieve these goals, we propose four aims. In Aim 1, we will develop and validate a novel data-driven computational framework for identifying neurobiological features that distinguish between groups, leveraging recent advances in deep learning and brain circuit dynamics. In Aim 2, we will identify neurobiological features that distinguish idiopathic psychosis (schizophrenia) from neurotypical controls, using our validated computational framework and “Big” data from schizophrenia. In Aim 3, we will identify neurobiological features that distinguish Alzheimer’s disease patients with psychosis from Alzheimer’s disease patients without psychosis, using our validated computational framework and data from Alzheimer’s disease and schizophrenia. In Aim 4, we will identify neurobiological features that predict onset of psychosis in Alzheimer’s disease. The proposed studies are highly synergistic with the goals of the PAR-20-159, which aims to “enhance knowledge of mechanisms associated with neuropsychiatric symptoms in persons with Alzheimer’s disease”. Through the successful completion of the work described here, the proposed studies will transform our understanding of brain circuit mechanisms underlying psychosis in Alzheimer’s disease, and crucially, provide a new computational framework for improved mechanistic understanding of other neuropsychiatric symptoms in Alzheimer’s disease. Ultimately, these advances will lead to better diagnosis and more effective treatments for neuropsychiatric symptoms in Alzheimer’s disease and, more broadly, advance precision medicine.

精神病症状是阿尔茨海默氏症中最常见和持续的神经精神症状之一 疾病，影响超过50％的阿尔茨海默氏病患者。然而，阿尔茨海默氏症的精神病病因 疾病仍然很少了解，系统的调查受到小样本的阻碍， 缺乏可再现的发现。至关重要的是，需要强大的生物标志物来了解 阿尔茨海默氏病的精神病，并确定治疗靶标。新近获得的人类神经生物学 数据为开发可用于精神病的强大和预测生物标志物提供了前所未有的机会 阿尔茨海默氏病。我们提案的总体目标是确定可靠和预测性的生物标志物 阿尔茨海默氏病的精神病使用了新型数据驱动的计算框架。具体来说，我们将使用 一种变革性的“大数据”科学方法，结合了令人兴奋的深度学习和我们的最新进展 关于定量动态大脑电路分析的最新工作，并进行了大量新的大规模开放的研究 来自多个财团的源脑成像和表型数据，以及我们在 斯坦福大学。为了实现这些目标，我们提出了四个目标。在AIM 1中，我们将开发和验证 新型数据驱动的计算框架，用于识别区分神经生物学特征 小组，利用深度学习和脑电路动态的最新进展。在AIM 2中，我们将确定 使用特发性精神病（精神分裂症）与神经型控制的神经生物学特征，使用 我们验证的计算框架和精神分裂症的“大”数据。在AIM 3中，我们将确定 神经生物学特征将阿尔茨海默氏病的精神病患者与阿尔茨海默氏病区分开 没有精神病的患者，使用我们经过验证的计算框架和来自阿尔茨海默氏病的数据 和精神分裂症。在AIM 4中，我们将确定可预测精神病发作的神经生物学特征 阿尔茨海默氏病。拟议的研究与Par-20-159的目标高度协同作用， 旨在“增强对与神经精神症状相关的机制的了解 阿尔茨海默氏病”。通过成功完成此处描述的工作，拟议的研究将 改变我们对阿尔茨海默氏病精神病的脑电路机制的理解，并 至关重要的是，提供了一个新的计算框架，以改善对其他的机械理解 阿尔茨海默氏病的神经精神症状。最终，这些进步将导致更好的诊断和 对阿尔茨海默氏病的神经精神症状的更有效治疗 精密医学。