情节记忆是人类生存和发展最为重要的基本能力之一，同时也是认知神经科学的基础前沿领域。情节记忆的一个最核心的特征是模式分离，它是指大脑能降低模式类似的记忆的相似性，从而降低不同记忆之间的竞争、干扰和混淆，从而促进新的学习。以健康老年人、轻度认知障碍（MCI）及阿尔海默茨病(AD)等为被试的研究发现模式分离能力在记忆老化和障碍的预测上有重要的作用。动物的研究结果解释模式分离主要由海马负责，但目前以人类为被试的研究还很少，特别是海马内部复杂精细结构如何支持模式分离的还不是很清楚。本研究采用经典的模式分离行为任务（BPS-O任务）和高清磁共振成像T1结构像来考察模式分离的神经结构基础。共有155名健康成年人被试（其中男性69名，女性86名，平均年龄22.7岁）参与了本实验。在学习阶段，被试需要对128张不同的物体图片做出“室内物”或“室外物”的判断。在测验阶段，有三种不同类型的图片，伪随机混合出现，被试需要判断每个图片做出新、旧或者相似的判断。这三种图片包括，完全相同的物体图片，被试应判断为旧；完全不同的新物体图片，被试应判断为新；与学习阶段材料类似的物体图片，但在颜色、方位、朝向等特征上发生了一些改变，被试应判断为相似。通过考察被试对不同相似程度图片的反应准确性，可以得到被试的模式分离AUC分数。利用基于体素的形态学分析方法VBM，并通过FreeSurfer软件对海马进行分割，探索再认记忆与模式分离的大脑结构基础。多体素模式分析发现右侧海马(hippocampus)和外侧枕叶(lateral occipital cortex，LOC)等脑区可以预测个体模式分离AUC分数；右侧海马和左脑旁海马(para-Hippocampus，PHA)以及双侧的颞中回(middle temporal gyrus，MTG)、额极(frontal pole，FP)、外侧枕叶(LOC)等脑区均可以预测被试个体再认记忆成绩；联合分析揭示右脑海马、外侧枕叶（right LOC）、双侧额极、以及左脑后部颞中回同时可以预测再认记忆以及模式分离水平。为了弥补多元模式分析无法区分海马精细复杂结构的缺陷，我们又使用FreeSurfer对海马子区进行了分割，结果发现右侧海马下托区体积与被试再认记忆显著负相关，左侧海马CA4和DG区体积之和与被试模式分离AUC分数显著正相关。本研究首次利用多体素模式分析对模式分离和再认记忆的大脑神经结构基础进行了全脑范围的探索，发现二者不仅局限于海马还涉及到如外侧枕叶等其他脑区；此外，还首次揭示了模式分离与再认记忆在海马子区结构基础上的分离；为促进模式分离认知神经机制的理解以及将来阿尔海茨默等老化障碍的临床诊断、预测奠定了基础。

Episodic memory is one of the most importantly primary ability for human survival and development, meanwhile it is the frontier fields of cognitive neuroscience. The term of pattern separation means that our brain makes the input information representation of similar stimulus to be more orthogonal, and reduces the similarity of the output information representation, consequently decreasing the competition and interference between old and new material, lessening the confusion of memory and facilitating the new study. Pattern separation is one of the most core characteristic of episodic memory. Some brain imaging studies which enrolled the participant including healthy elder, MCI and AD have showed the level of damage in pattern separation is correlated significantly with the injury of perforant pathway in hippocampus. Pattern separation is more sensitive and effective than traditional neuropsychological test in clinical diagnosing the memory impairment of aging. Pattern separation not only has important scientific theoretical significance, but also has profound effect on clinic. Therefore, it has become the hot topic which attracting extensive attention of cognitive neuroscience. Previous related researches have demonstrated that hippocampus responsible for pattern separation mainly. But most of studies were animal studies, there was only little study on human and most of them focused on fMRI studies and DTI studies. There is no direct study which try to explore the neural basis of brain structure of pattern separation in human, especially the basis of complicated and fine brain structure in hippocampus. This study collected high-resolution magnetic resonance anatomical imaging of 155 healthy adult (86 female, 69 male; mean age, 22.7). We adopted the classic task of pattern separation (BPS-O Task) to get both the performance of pattern separation and recognition memory .Our study combined the data of brain structural and multi-voxel pattern analysis (MVPA) to explore the neural basis of brain structure of pattern separation in human. The results of MVPA showed that (1), the right hippocampus and LOC can predict the score of pattern separation AUC score, respectively; (2), the right hippocampus and LOC, MTG, left PHA can predict the performance of recognition memory; (3), conjunction analysis revealed right hippocampus, LOC, frontal pole and lp-MTG can predict both the performance of pattern separation and recognition memory. By surface-based analysis, we found that (1), the volume of CA4/DG, or CA2/CA3 in the left hippocampus both show a positive correlation with AUC Score; (2), there was a negative correlation between recognition memory and right subiculum.This is the first study which combines univariate surface/voxel-based analysis and MVPA to explore the neural basis of brain structure of pattern separation, meanwhile we use FreeSurfer to elaborate effect of subfields of hippocampus. Our study indicated that the pattern separation of objects in human is not only based on hippocampus but also involves the lateral occipital lobe. We found that the neural basis of brain structure of pattern separation and recognition memory exist the common area and the different area in human brain. This finding promotes the understanding of the neural mechanism of pattern separation and the precision of clinical diagnosis of AD, as well as other aging disorders.