专题链接

学术看板 当前位置:网站首页 > 专题链接 > 学术看板

学术讲座:美国卡内基梅隆大学李远宁博士讲座

作者:计算机学院 来源:计算机学院 阅读次数:229日期:2017/12/22

报告题目:Temporal dynamics of face and word processing in the human fusiform

报告时间:2017年12月27日(周三) 上午10:30-12:00

报告地点:清水河校区主楼B1-501会议室

报告人:李远宁 博  士    美国卡内基梅隆大学

主持人:顾  实 教  授    计算机科学与工程学院

 

内容介绍:The fusiform gyrus contains category-selective patches that are critical for visual recognition, with damage to these patches leading to category-selective impairments in object recognition, such as acquired alexia and prosopagnosia. However, many gaps remain in our understanding of the dynamic role that fusiform plays in contributing to multiple stages of category-specific information processing. To assess the information processing dynamics of the human fusiform, here we report results from 8 subjects with intracranial electrodes placed directly on word selective (4 subjects) or face selective (4 subjects) fusiform gyrus (the "visual word form area [VWFA]" and "fusiform face area [FFA]" respectively). Specifically, we use multivariate pattern recognition methods in conjunction with multiple face and word processing paradigms to uncover common neurodynamic information processing principles of category-selective fusiform gyrus. The results show strong decoding accuracy for faces and words in the FFA and VWFA respectively, first becoming statistically significant between 50-100 ms and peaking between 150-250 ms after stimulus onset. Next we examined the dynamics of within category decoding. For words significant decoding was seen in both subjects between approximately 150-300 ms wherein visually similar words could not be decoded from one another, but dissimilar words could be decoded (organized by orthographic similarity). There was a later phase between approximately 250-500 ms where even orthographically similar words could be significantly decoded from one another (individual-level representation). For faces significant expression-invariant decoding was seen in each subject in the same 250-500 ms time frame. The neural response for faces was organized by facial feature similarity, emphasizing the role of the eyes and mouth in face individuation. Taken together, these results suggest a multi-stage information processing dynamic wherein the representation in category-selective fusiform gyrus evolves from a coarse category-level representation to an invariant and highly detailed individual representation over the course of 500 ms.

 

主讲人介绍:李远宁,卡内基梅隆大学(Carnegie Mellon University)计算神经科学与机器学习专业在读博士。2011年本科毕业于北京航空航天大学高等工程学院电子信息工程专业,2013年在美国卡内基梅隆大学获得电子与计算机工程硕士学位。主要利用机器学习和高维统计推断结合人脑介入式电极的方法研究大脑高级视觉皮层物体识别的神经原理,从事拓展大脑物体识别区域尤其是面孔和文字区域的信息处理的时间和空间特性相关的研究。系列研究工作获得美国国立卫生研究院(National Health Institute)和国家科学基金(National Science Foundation)的基金支持,研究成果发表于美国科学院院刊(PNAS),自然-通讯(Nature Communications),神经影像(Neuroimage),Cortex等期刊上,并曾被国际认知神经科学学会(Cognitive Neurosciene Society)主页等报道。