Identifying Patterns of Cognitive, Motor, and Brain Structural Abnormalities Differentiating Alcohol Use Disorder with and without HIV Infection Comorbidity

摘要 酒精使用障碍(AUD)发生在美国最大一部分人中，他们是 依赖于一种物质。AUD在个体和人类中的共同发生 免疫缺陷病毒(HIV)感染率很高，发病率是美国的两倍 普通人口。AUD和HIV感染各自导致大脑结构的破坏 正直与认知和运动障碍，影响一些不同的和一些重叠的 神经系统，但也会加剧对选择性系统的不良影响 协同或相加工艺。这个研究项目的目标是开发新的机器 辨证这两种疾病的构成因素和作用的学习方法 让大脑参与提高对HIV/AUD共病的机制和动态的理解 对大脑的影响。 与AUD和HIV共病相关的不良反应的有效研究 在脑形态上需要识别多个诊断分组之间的差异，即， 健康对照(CTRL)、HIV阴性酗酒者(AUD)、HIV阳性而不饮酒 依赖(艾滋病毒)和艾滋病毒阳性酒精依赖(艾滋病毒/澳州)。测试推论 对复杂疾病的多个诊断分组通常会产生不确定的或 由传统的单变量横断面研究设计得出的相互矛盾的结果， 被构造为一次对两个队列进行建模，并将“讨厌”变量保持不变 包括多种因素的权力，这些因素包括很可能是 与区分原发疾病有关。 在这里，我建议开发健壮和可靠的机器学习技术，可以 从每个队列独有的纵向MRI数据中确定模式。这种类型的数据驱动 方法能够同时分析所有数据点，从而提供替代方法 与专注于由专家领域指导的数据子集关联的传统方法不同 知识。我设想这个项目将扩大我对神经精神病学的理解。 AUD和HIV，也将导致扩展对神经因素的机械论理解 与HIV/AUD共病有关。 提高对HIV/AUD共病和诊断差异的机制认识 可以提高医生和护理者的意识，并帮助临床医生开发有针对性的 持续症状受益的治疗选择。

SUMMARY Alcohol Use Disorder (AUD) occurs in the largest segment of individuals in the U.S. who are dependent on a substance. The co-occurrence of AUD in individuals with human immunodeficiency virus (HIV) infection is high, occurring at twice the rate as occurs in the general population. AUD and HIV infection each are responsible for disruption of brain structural integrity and cognitive and motor impairments, affect some different and some overlapping neural systems, but can also exacerbate the untoward effects on selective systems through synergistic or additive processes. The goal of this research project is to develop novel machine learning methods to differentiate the compounding factors and effects of these two disorders involving brain to improve the mechanistic and dynamic understanding of HIV/AUD comorbidity effects on the brain. Efficient study of the untoward effects associated with the comorbidity of AUD and HIV on brain morphology requires identifying differences across multiple diagnostic groupings, i.e., healthy controls (CTRL), HIV-negative alcoholics (AUD), HIV-positive without alcohol dependency (HIV), and HIV-positive with alcohol dependency (HIV/AUD). Testing inferences across multiple diagnostic groupings of complex disorders commonly yields inconclusive or conflicting findings when done by conventional, univariate, cross-sectional study designs, which are constructed to model two cohorts at a time and hold "nuisance" variables constant with little power to include multiple factors comprising the complexity and dynamism that may well be relevant for distinguishing the primary disorders. Herein, I propose to develop robust and reliable machine learning technology that can identify patterns from longitudinal MRI data unique to each cohort. This type of data-driven method is able to analyze all data points concurrently and thus provide alternative approaches to conventional methods, which focus on correlating subsets of data guided by expert domain knowledge. I envision that this project will expand my understanding of the neuropsychiatry of AUD and HIV and will also result in extending a mechanistic understanding of neurofactors relevant to HIV/AUD comorbidity. Improving the mechanistic understanding of HIV/AUD comorbidity and diagnostic differences may enhance physician and caregiver awareness and aid clinicians in developing targeted therapeutic options for sustained symptomatic benefits.