Systems-based machine learning approach to understanding clinical, genetic, and pathophysiological heterogeneity in Parkinson's dementia

简要说明研究的背景，包括潜在的影响帕金森病(PD)是高度异质性的。虽然帕金森病是第二常见的神经退行性疾病，但我们对其进展的临床和神经解剖学序列的了解有限。一半的确诊患者也会在10年内患上痴呆症，这些患者的预后通常更差。此外，还没有发现可以预测帕金森氏症发病风险或跟踪疾病进展的可靠生物标志物。大型多模式数据集的可获得性，如帕金森进展标记倡议和帕金森研究中的UCL愿景，为部署数据驱动的疾病进展建模和分型技术的最新发展提供了机会，以加深我们对帕金森病的理解。随后，这项研究可能会为帕金森病患者的临床试验招募提供信息，这些人患痴呆症的风险更高。目的和目的通过结合来自多模式神经成像、视网膜成像和临床评估(包括新的和传统的)的可用的纵向数据，总体目标是描述帕金森痴呆的临床、遗传和病理生理的异质性。具体地说：1.使用数据驱动的疾病进展建模，根据多模式数据集中可观察到的异常的顺序、时间和严重程度来确定PD进展亚型。2.通过仅使用神经成像数据重复亚型实验，识别PD进展的医学成像特征。同样，对于新的视力测试，以期了解帕金森病痴呆的进展情况。3.通过量化每个亚型的特征重要性，确定PD个性化药物所需的最小特征集。因此，将PD痴呆亚型与临床症状和病理生理联系起来。研究方法的新颖性本研究的主要新颖性是从横断面数据中提取纵向信息的能力，以表征和确定PD的亚型。具体地说，使用了子类型和阶段推理算法，这是一种机器学习技术，可以提取具有特定时间序列模式的表型。PD特征(序列、时间和症状严重性)的联合估计将使用单个统一模型进行，该模型解决了数据的相对稀疏性，但利用了深度。帕金森氏病的视觉研究还包含新的神经成像技术，以检测大脑连接的早期变化，例如基于Fixel的白质完整性分析和定量易感性图谱。与EPSRC的战略和研究领域保持一致本项目与EPSRC的几个战略和研究领域保持一致。然而，主要是与“用于传感和分析的颠覆性技术”保持一致，因为一种新的方法正在被应用于进一步了解PD并确定任何特定于亚型的潜在生物标志物。根据项目的进展，这可能会转化为临床支持系统(因此可能会“改变社区护理”)。具体地说，该项目旨在为临床试验招募提供信息，从而可能导致“未来疗法的发展”和“治疗优化”。该项目与临床医生(二级主管)密切合作，为项目设计提供信息。关于EPSRC在医学成像领域的战略重点，该项目解决了两个高度优先的领域：-使更早和更有效地诊断身体和精神健康状况，为治疗计划提供信息-从临床数据/图像整合现有和更多的信息(例如，通过机器学习和/或数学科学技术)任何公司或合作者参与该项目没有签署的公司或合作者。

Brief description of the context of the research including potential impactParkinson's disease (PD) is highly heterogeneous. Although PD is the 2nd most common neurodegenerative disease, we have a limited understanding of the clinical and neuroanatomical sequence of its progression. Half of diagnosed patients also develop dementia within 10 years and these patients typically have a worse prognosis. Additionally, no robust biomarkers to predict risk of developing Parkinson's, nor to track disease progression have been identified. The availability of large multimodal datasets, such as the Parkinson's Progression Markers Initiative and the UCL Vision in Parkinson's study, presents an opportunity to deploy the latest developments in data-driven disease progression modelling and subtyping technology to deepen our understanding of PD. Subsequently, this research may inform clinical trial recruitment for people with PD at higher risk of developing dementia. Aims and ObjectivesThe overarching objectives is to characterise the clinical, genetic, and pathophysiological heterogeneity in Parkinson's dementia by combining available longitudinal data from multimodal neuroimaging, retinal imaging, and clinical assessments (both novel and traditional). Specifically: 1. Identify PD progression subtypes in terms of the sequence, timing, and severity of observable abnormality in multimodal datasets using data-driven disease progression modelling. 2. Identify medical imaging signatures of PD progression by repeating subtype experiments using only neuroimaging data. Likewise, for novel tests of vision, with a view to understanding progression in PD dementia. 3. Identify the minimum set of features necessary for personalised medicine in PD by quantifying feature importance for each subtypeThus, to link PD dementia subtypes to clinical symptoms and pathophysiology.Novelty of Research MethodologyThe main novelty of this research is the ability to extract longitudinal information from cross-sectional data, in order to characterize and determine subtypes of PD. Specifically, the Subtype and Stage Inference algorithm is used, which is a machine-learning technique that can extract phenotypes with specific temporal progression patterns. Joint estimation of characteristics of PD (sequence, timing, and symptom severity) will be performed using a single unified model, which tackles the relative sparseness of the data, but takes advantage of the depth. The Vision in Parkinson's disease study also contains novel neuroimaging techniques to detect early changes in brain connectivity, such as fixel based analyses of white matter integrity and quantitative susceptibility mapping. Alignment to EPSRC's strategies and research areasThis project aligns to several of the EPSRC's strategies and research areas. However, the main alignment is with "Disruptive technologies for sensing and analysis", as a novel method is being applied to further understand PD and determine any potential biomarkers specific to subtypes. Depending on the progress of the project, this may potentially translate to a clinical support system (and thus may "transform community care"). Specifically, the project aims to inform clinical trial recruitment and thus may lead to the "development of future therapies" and "treatment optimisation". This project works closely with clinicians (the secondary supervisor) to inform project design. With regards to the EPSRC's strategic focus within Medical imaging, this project tackles two areas of high priority: - Enabling earlier and more effective diagnosis of physical and mental health conditions, to inform treatment planning- Integration of existing and additional information from clinical data/images (e.g. via machine learning and/or mathematical science techniques)Any companies or collaborators involvedThere are no signed companies or collaborators in this project.