Unimodal, multimodal and machine-learning techniques to identifying structural, functional and connectivity dynamics underlying empathic accuracy

单模态、多模态和机器学习技术，用于识别共情准确性背后的结构、功能和连接动态

• 批准号: RGPIN-2020-06964
• 负责人: Shane, Matthew
• 金额: $ 2.04万
• 依托单位: University of Ontario Institute of Technology
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=765607
• 关键词: Unimodal; multimodal; machine; learning; techniques

The ability to intuit another's feeling states is a critical component of human interaction. This ability is believed to rely on both perspective-taking (PT) and empathic concern (EC), which together may represent the ability, and motivation, to consider another's point of view, respectively (Keysers Zaki, 2009; Arbuckle Arbuckle & Shane, 2016). However, this research remains nascent. First, most work to date has evaluated neural activity while participants attempt to understand another's mental state, with only a handful of studies evaluating neural circuits during the successful understanding of those mental states (referred to as `empathic accuracy' (eACC)). Second, most work to date has reported only single imaging modalities (ie. brain structure, function or connectivity), that cannot document potentially important multimodal patterns. Thus, little is currently known about cross-modal brain dynamics that support the accurate understanding of another's feeling states. The present proposal describes our lab's next intended projects in this space, aimed at targeting these specific limitations. Aim 1 involves use of standard univariate methods to evaluate structural (T1-T2-sequences), functional (fMRI), and resting state functional connectivity data, to identify complex patterns underlying PT, EC and eACC. Aim 2 involves the use of joint Independent Component Analysis (jICA) to "fuse" all three modalities into a single data matrix, to allow for full consideration of cross-modal predictors of eACC. Finally, Aim 3 involves use of machine-learning techniques to construct and train a multimodal predictive model of eACC, and to test generalizability of that model in two independent, out-of-sample datasets. These studies will fill important gaps in knowledge, towards our lab's long-term goals of delineating the neural mechanisms underlying representation of other's thoughts/feelings. Results from these studies will be presented at international conferences and published in high-tier academic journals, and will be of considerable interest to academics studying the neural underpinnings of cognitive and emotional processes. They will also contribute to student training at both the undergraduate and graduate level, by providing exposure and training in collection, analysis, interpretation and ethics of a large-scale multimodal neuroimaging study. These students, diversely trained in psychology, neuroscience, and sophisticated analytical and machine learning methods, will serve as the next generation of Canadian researchers, thereby contributing directly to NSERC's stated mission of diversifying and energizing a new community of future NSE researchers.

凭直觉感知他人情感状态的能力是人类互动的重要组成部分。这种能力被认为依赖于观点采择（PT）和同理心关注（EC），它们共同代表了考虑他人观点的能力和动机，分别（Keysers Zaki，2009;阿巴克尔阿巴克尔& Shane，2016）。然而，这项研究仍处于初期阶段。首先，迄今为止，大多数研究都是在参与者试图理解另一个人的精神状态时评估神经活动，只有少数研究在成功理解这些精神状态时评估神经回路（称为“移情准确性”（eACC））。其次，迄今为止，大多数工作仅报告了单一成像模式（即。大脑结构、功能或连通性），不能记录潜在重要的多模态模式。因此，目前对支持准确理解他人感觉状态的跨模态大脑动力学知之甚少。 本提案描述了我们实验室在这一领域的下一个预期项目，旨在针对这些特定的限制。目的1涉及使用标准的单变量方法来评估结构（T1-T2-序列），功能（fMRI）和静息状态功能连接数据，以识别PT，EC和eACC的复杂模式。目标2涉及使用联合独立成分分析（jICA）将所有三种模式“融合”到一个单一的数据矩阵中，以充分考虑eACC的跨模式预测因子。最后，目标3涉及使用机器学习技术来构建和训练eACC的多模态预测模型，并在两个独立的样本外数据集中测试该模型的可推广性。这些研究将填补重要的知识空白，实现我们实验室的长期目标，即描绘表达他人思想/感受的神经机制。 这些研究的结果将在国际会议上发表，并发表在高级学术期刊上，对研究认知和情感过程的神经基础的学者来说，将产生相当大的兴趣。他们还将通过提供大规模多模态神经成像研究的收集，分析，解释和道德方面的暴露和培训，为本科和研究生阶段的学生培训做出贡献。这些学生，在心理学，神经科学和复杂的分析和机器学习方法的训练，将作为下一代的加拿大研究人员，从而直接有助于NSERC的多样化和激励未来NSE研究人员的新社区的既定使命。