海量磁共振全脑影像数据的深度挖掘分析

磁共振成像是开展脑科学研究最重要的现代影像方法之一。随着近几年各国脑计划的陆续展开如美国人类连接组计划 (Human Connectome Project) 和 2013 年美国奥巴马政府制定的“脑计划” (BRAIN Initiative)，磁共振技术已被广泛应用于探索大脑结构和功能网络连接图谱。使用脑网络联接图谱研究情绪障碍类疾病，一般是通过统计性比较正常人和病患的脑网络，寻找有意义的差异特征，为疾病的诊断和治疗提供指导。但是，高分辨的全脑磁共振成像经常产生大量的结构和功能联接图谱数据，因此需要非常有效的计算技术来实现这一过程。我们计划首先建立人-猴相对应的脑疾病影像数据集合，发展适用于神经信息分析的机器学习算法，模拟分析脑网络大数据，并结合模式动物和人类临床实验的证据，探索疾病机理，发展有效的治疗手段。

Deep Mining and Analysis for Massive Data of Whole Brain Magnetic Resonance Imaging

Magnetic resonance imaging is one of the most important modern brain imaging methods to carry out biomedicine research. Recently in several national level brain research projects such as the Human Connectome Project and the BRAIN Initiative, magnetic resonance imaging technique has been widely used to acquire whole-brain structure and function maps of network connections. These brain-mapping studies, generally through statistical group comparison between healthy subjects and clinical patients, allow to seek significant pathological differences for providing better guidance for the diagnosis and treatment of mental diseases. However, high-resolution magnetic resonance imaging of the whole brain often produce a large amount of structural and functional connectomics data, demanding effective computing technology to optimize this procedure. We plan first to collect cross-species magnetic resonance imaging data in both human and monkey brains, to develop novel machine learning algorithms for neural information analysis, and to combine evidence of animal models and human clinical trials for better understanding of the pathological mechanisms and for the development of effective therapeutic interventions.