物體辨識與選擇於人類大腦神經表徵的空間與時間動態性研究

Abstract

個體進行物體辨識與選擇高度仰賴大腦中對該物體神經表徵的活化，以利處理視覺資訊並與外在環境互動。過去研究已發展多種多變量型態分析方法探討物體辨識與選擇在人類視覺皮質裡的功能性機制與神經表徵資訊。然而，是否能夠以及如何整合類別特定腦區中物體神經型態的空間與時間特性仍不清楚。本論文藉由多變量神經型態分析方法測量物體辨識與選擇神經表徵的時空資訊與其動態性。在系列研究一裡，功能性磁振造影的多變量神經型態結果顯示類別選擇腦區與物體相關腦區分別展現人臉與中文文字倒立效應的獨特與共享神經表徵型態。在系列研究二中，我們使用腦磁圖儀進一步測量人臉與中文文字倒立效應的神經表徵在時間上的動態性。透過時間廣義化分析，我們可以區分神經表徵的正倒立資訊與類別資訊，並展現神經表徵的時間動態性。最後，在系列研究三，我們檢驗了上而下注意力調節對物體選擇的功能運作與其神經型態的時空動態。透過功能性磁振造影與腦磁圖儀的融合分析，我們展現了人臉選擇腦區與景物選擇腦區的神經空間與動態特性。值得注意的是，我們類別選擇腦區神經型態的表徵相似性調節效果的來源腦區為前額葉。綜合上述的發現與研究規劃，透過整合不同的多變量神經型態方法探究物體辨識與選擇的神經表徵，將可以增進瞭解個體物體知覺辨識與記憶再認歷程神經表徵在空間與時間上的動態性與互動性，並釐清物體知覺與短期記憶歷程的神經機制與心理歷程。

Object recognition and selectivity rely heavily on the activation of neural representations for the objects in the human brain, enabling individuals to process and interact with external information within real-world environments. Previous studies have developed multivariate pattern analyses to investigate the functional mechanisms and neural representations underlying object recognition and selectivity in the human visual cortex. However, it remains unclear whether and how spatial and temporal information of object recognition and selectivity can be represented within the category-selective brain regions and integrated across different imaging modalities. In this work, we adopt the multivariate pattern analysis approach to investigate the spatiotemporal dynamics of neural representations for object recognition and selectivity. In Study I, our fMRI results using multivariate pattern analysis demonstrated the distinct and shared configural representations for the inversion effects for face and Chinese character recognition in category-selective and object-related brain areas. In Study II, we further tracked the temporal dynamics of the inversion-related neural responses for face and character recognition using MEG with the multivariate decoding approach. Through temporal generalization analysis, we decoded the neural information about stimulus orientation and category over and across time for faces and characters. In the final study, we examined the spatiotemporal profiles of the functional operations and the underlying mechanisms through the top-down modulation of attention on object selectivity in both face-selective and scene-selective brain regions using an fMRI-MEG fusion approach. The fusion results showed the attentional enhancement of representational similarity in category-selective brain areas. Importantly, we also observed that the goal-directed modulation of representational similarity in the category-selective brain areas was guided by the prefrontal cortex using the Granger causality analysis. Together, this work provides novel evidence improving our understanding of the spatiotemporal dynamics of neural representations for object recognition and selectivity in the human brain.