基于多模态磁共振影像特征的个体判别与个体预测，是神经影像计算领域的重点研究方向。研究表明，多模态磁共振影像可以用来研究活体人脑连接与人脑网络，该领域已成为神经科学、神经精神病学、及神经影像学的研究热点之一。最近，越来越多的磁共振影像个体判别/预测研究已经开始尝试采用磁共振脑连接或脑网络信息作为分类/预测特征，并取得了不错的效果。但到目前为止，领域内仍缺乏对相关计算方法的系统深入研究，这导致了在很多关键计算处理步骤的选择上缺乏依据，给相关应用研究造成了很大困扰。针对这一困境，本项目拟从三个方面开展深入系统的相关计算方法学研究：１）不同脑网络原始特征的评价与优化；２）海量脑网络特征的降维（即特征选择）及融合方法研究；３）个体判别/预测方法及其可推广性研究。最后，为了进一步验证和应用相关的方法学研究成果，将开展４）围绕人类阅读能力的个体判别（即阅读障碍个体）与个体预测（即阅读认知分数）应用研究。

Individual discrimination and prediction using multi-modal MRI features has been a critical topic in the field of neuroimaging computation. During last decade, multiple MRI techniques have been widely applied to non-invasively study human brain connectivity and networks in-vivo. These MRI-based connectivity/network studies have attracted a lot of attention in neuroscience, neurology, psychiatry, and neuroimaging...Recently, there are increasing studies applying the MRI-based connectivity/network features for individual discrimination and prediction. However, a number of methodological issues remain unsolved, hindering relevant applications. Our project is proposed to exclusively address these issues. Specifically, our investigation will include 1) accessing/optimizing the extract methods of raw MRI-based connectivity/network features; 2) analyzing the methodology of feature selection and fusion of MRI-based connectivity/network features; 3) evaluating/developing the discrimination/prediction models as well as their generalization abilities. Based on these relevant methodological results/conclusions, we will try to apply MRI-based connectivity/network features to achieve accurate discrimination for dyslexic individuals, as well as accurate prediction of individual scores of reading skills within healthy populations.