Abstract
摘要:獼猴的初級視覺皮層V1,為瞭解人類的早期視覺系統提供了一個非常好的實驗模型。根據理論模型的結果,V1被認為是一個動態的非線性系統,V1有能力去快速適應不同視覺場景而做出相對應的反應。這個概念跟傳統對V1的觀點是很不一樣的:傳統觀點認為V1是由一些線性和非線性具有固定特質的濾波器所組成。本研究的第一個目標就是設計來測試這兩個對V1不同的概念。此外,根據解剖和神經生理學的研究,V1不再被視為是一個單一的神經網絡,而是由幾個鬆散且互聯堆棧的皮質層所組成的複雜神經網路。每個皮質層從不同的結構接收訊號,並將訊號輸出到不同的結構。本研究的第二個目標是研究視覺信息是如何在不同的V1皮質層做處理及轉譯。目標1:空間視覺接區域在不同的V1皮質層。我們將用四個不同型態的視覺刺激,測量不同V1皮層中神經元的視覺接區域。這些刺激包括:自然風光,白噪聲,稀疏噪音,和Hartleys(光柵)。第一個目標是比較從這四個刺激所獲得的空間視覺接區域。假如V1是由最簡單的神經網絡組成,這四個刺激所獲得的空間視覺接區域應該很類似,然而初步結果卻顯示很顯著的不同。這個研究的主要目的是要揭示什麼刺激參數,會造成V1強烈的情境效果。目標2:顏色的空間視覺接區域。V1中有很多的神經元都會對顏色產生反應。第二個研究我們將用不同顏色屬性的光(相當於長、中、短錐狀細胞所敏感的光波長)及視覺刺激,去測量V1的空間視覺接區域。我們也將對神經元在純色對比的視覺刺激下,測量其對不同方向性及空間頻率的光柵、及因不同顏色所產生的邊緣刺激的反應。這個目標將有助於釐清及解釋來自不同實驗室運用不同技術所獲得的不同結果。目標3:非線性動態反應在不同的V1皮質層。一些在V1 2/3皮質層的神經元會對 100 毫秒的對比刺激產生非線性的動態反應。在此我針對這個現象提出了一個更全面性的研究。我們將用不同型態的視覺刺激及之前所用對比刺激,測量在V1不同皮質層的神經元的非線性動態反應。這個結果將有助於回答這些問題:“什麼是一個簡單的細胞?”和“什麼細胞是簡單的?”總體而言,這項研究的目的是要了解V1的神經迴路以及V1神經元是如何處理的明暗和色彩的視覺訊息。了解V1的神經網絡將對腦皮質組織處理訊息的一般規則提供寶貴洞見,並有助於V1電腦神經網絡的建立。
Abstract: The macaque monkey’s early visual system up to primary visual cortex V1 is a good model for the human early visual system, which is why it has been the focus of a lot of vision research. Theoretical models suggest that V1 is a dynamical nonlinear system leading to the concept that V1 has the capacity for rapid adaptability to the statistics of the visual scene. This concept is different from the more traditional view that V1 can be viewed as a bank of linear and nonlinear filters of visual information. This first goal of the research proposal was designed to test the two different concepts. Moreover, based on anatomy and neurophysiology our view of V1 has changed. Instead of viewing V1 as a single network, we now conceive of the cortical layers of V1 as a stack loosely interconnected but distinct neuronal networks. Each lamina has different specific inputs, projecting targets and feedback connections. The second goal of the research proposal is to investigate how visual information is processed across different V1 layers. Aim 1: Spatial receptive fields in different layers of monkey V1. We will measure spatio-temporal maps of neurons throughout V1 layers with four different sets of stimuli: natural scene, white noise (m-sequences), sparse noise and Hartleys (gratings). The first goal is to compare the spatial maps obtained from the four stimulus sets because for the simplest kind of neural networks they should yield the same maps, and differences (as shown in Preliminary Results) will be revealing. From the spatio-temporal maps obtained with white noise, we will synthesize predicted responses to the contrasts steps. This aim was designed to reveal what stimulus parameter, such as spatial correlation or sparseness, may contribute to the strong contextual effect in V1. Aim 2: Spatial maps in color. A significant fraction of macaque V1 cells respond to color. Next we propose to study the color properties of V1 responses by measuring spatial maps of the L, M and S cones input to a cell, with m-sequences, sparse noise and Hartleys. We will compare the spatial maps for cells in different layers with orientation and spatial frequency tuning curves for gratings patches defined by pure color contrast, and with the orientation specificity for edges defined by color contrast. This aim is important because it may help reconcile divergent results obtained from different laboratories with different mapping techniques. Aim 3: A dynamic nonlinearity in V1 layers. A dynamic nonlinearity in responses to 100 msec pulses of contrast has been revealed in layer 2/3 cells. Here I propose a more comprehensive study of the responses to pulses and to other stimulus ensembles of cells in different V1 layers. This aim is significant because it will help answer the questions, “what is a simple cell?” and “what cells are simple”. Overall, the objectives of the study are to understand neuronal circuitry in V1 and how it processes brightness and color. Understanding V1 circuitry will offer invaluable insight about the general rules of cortical organization.
Keyword(s)
初級視覺皮層
視覺接區域
簡單細胞
primary visual cortex
receptive field
simple cell