基于解释卷积网络的运动-认知联合脑网络机制及认知量化研究

Large information exchanging resources in the brain are required under the execution of cognitive and motor joint tasks. It is necessary to study the movement cognition joint brain network in order to clarify the two information interaction pathways. Understanding the brain network can put forward more scientific evaluation mechanism for joint function to brain lesions such as frontotemporal lobar degeneration patients to provide medical advice. Previous works indicate cognitive and motor network theories are discussed independently, key technologies for cognitive dynamic information and supported experiments are still inadequate, there are still many problems on how to effectively provide for clinical cognitive quantification. This study proposed to use brain computer interface, the repeatable virtual reality test environment, dynamic brain network model and interpretable convolutional neural network algorithm to simulate the dynamic changes of patients with frontotemporal degeneration and normal brain network process, analytical quantification of pathological features can be distinguished based on the interpretation convolutional network to extract discriminated features and apply on the deconvolution neural network, therefore more comprehensive research on human brain cognitive-motor function. The ultimate goal is to understand the development of cognitive-motor brain network mechanism, to provide objective quantitative indicators for cognitive decline in frontotemporal lobar degeneration, and to provide strengthen guidance on the prevention of cognitive and motor decline.

人类大脑在完成复杂的认知或者运动任务时需要进行大量的信息交互，其信息环路的研究在理解大脑的工作机制和治愈大脑疾病中至关重要。前期研究发现部分认知功能区损伤患者除存在认知功能受损外，同时出现了运动功能障碍。这类病人不仅存在独立脑区功能缺失，也可能在功能区间信息交互通路上受损。课题拟针对认知对运动功能调控的认知-运动脑网络，基于联合功能受损的额颞叶变性患者，深入研究认知与运动脑功能区间通路和联合功能病变病理；依托脑机接口、深度学习与虚拟现实技术，利用可重复的虚拟现实测试环境，脑网络模型和解释卷积网络建立动态联合脑网络特征的评估模型，再结合解释网络提取特征和反卷积网络对可区分的病理特征进行解析量化为临床提供高可重复性认知-运动能力联合量化评价指标。本课题将为发展完善认知-运动联合脑网络机制提供新的科学依据，为额颞叶变性的认知衰退提供客观量化指标，在预防认知运动能力衰退及靶向强化指导上提供帮助。

阿兹海默症作为一种神经退行性疾病不仅症状在病情早期易与正常人老化混淆，而且病情发展不可逆，目前对其诊断方法仍未清晰。脑电(Electroencephalogram, EEG)可以实时有效地反映大脑活动的变化，其活动特征的分析可能用于阿兹海默症的诊断。根据患者和健康组每天认知训练的对照数据，通过多维度脑电特征分析，探究使用脑电特征区分健康人与病人的可能性，并探究通过认知训练的方式在治疗干预上是否有足够效果。根据不同脑电特征维度的事件相关电位时域分析、傅里叶变换计算功率谱频域分析和相位锁定值的脑功能网络分析结果，证明了使用脑电分析区分健康老人与早 期阿兹海默症病人是可行的。