Sex, function and structure: machine learning the human cerebral vasculature.

点击翻译按钮获取中文摘要

Brain activity is dependent on an adequate blood flow, supplied through a complex and difficult to measure vasculature. Because of technological limitations, and time constraints researchers typically investigate cerebral blood flow in one dimension; quantifying either function (time) or structure (spatial) in adult males.Thus, in the few studies stratifying sex differences, it is difficult to discern why women have a higher cerebral blood flow than men, but have divergent incidence's of cerebrovascular disease (stroke, dementia, Alzheimer's disease). Vascular ultrasound (US) measures forces impacted by physiological changes over time. Whereas, magnetic resonance imaging (MRI) provides high spatial resolution quantification of artery structure. Machine learning can enhance cerebrovascular assessment by multiplying valid high-fidelity signals from traditional techniques and reduce time constraints. Despite these techniques, there is limited research investigating cerebrovascular sex differences, and no data directly comparing cerebral artery function and structure in men and women using a combination of these approaches. Thus, this research aims to develop approaches that enhance the measurement of the cerebral vasculature in humans. This will be achieved through three strategic research aims: Aim 1: Develop, collect, and compare measures of cerebrovascular structure and function in men and women using high temporal (US) and spatial (MRI) resolution imaging: Our lab specializes in measuring physiological function of cerebral arteries, using vascular US, and has access to a research designated MRI. This proposal will utilize a battery of tests to assess sex differences in the cerebral vasculature using intra-class correlation and co variate analysis to describe structure and function metrics. Aim 2: Independent and supervised machine learning approaches will be used to stratify the mechanisms of brain blood flow regulation in humans: Machine learning algorithms will generate thousands of unique cerebrovascular outcomes from signals collected in Aim 1. Independent machine learning, through fusion algorithms, will develop a comprehensive index of cerebrovascular function by incorporating blood pressure, metabolism, respiratory and neural responses. We hypothesize that machine learning will enhance cerebrovascular assessment compared to traditional uni-dimensional analysis. Aim 3: Combine US and MRI measures of the cerebral vasculature to quantify the relationship between structure and function in humans: A tertiary focus of this research is the integration of state-of-the-art 3-dimensional vascular modelling of the major cerebrovascular structures acquired from MRI images, with our multi-factorial measures created in Aim 2. Integration of MRI and US measures will enable the quantification of abnormal flow patterns in multiple cerebrovascular vessels simultaneous, respective risk to cerebrovascular function during physiological stimulation.