Lecture Notes: Primary Visual Cortex and Cortical Modules PDF

Primary Visual Cortex and Cortical Modules How is visual information processed? A. How do you recognize this as a face? 1. Visual processing occurs in stages. 2. Retina processing followed by V1 processing followed by V2, V3 processing.... B. In V1 (primary visual cortex) how is information analyzed...

Primary Visual Cortex and Cortical Modules How is visual information processed? A. How do you recognize this as a face? 1. Visual processing occurs in stages. 2. Retina processing followed by V1 processing followed by V2, V3 processing.... B. In V1 (primary visual cortex) how is information analyzed? 1. The visual field is broken up into approximately 1000 parts and information from each location is analyzed by neurons in ~ 1000 different cortical modules. The diagram How the visual field is subdivided into regions in primary visual cortex + Each square area in primary visual cortex above represents one of the ~ 1000 cortical modules. The purple area in the visual field would be analyzed by the purple cortical module and all the neurons in that cortical module would respond only to light from the purple area in the visual field. © Hongdian Yang. This content is protected and may not be shared, uploaded, or distributed. below shows both the visuotopic/retinotopic map in V1 and how it is parceled into cortical modules: Cortical modules A. Cortical modules are hypothetical constructs but they have a physiological basis. Conceptually, each cortical module in primary visual cortex can be thought of as a computer containing all of the neural machinery needed for the initial analysis of the visual information in a single location in the visual field. Different neurons in each cortical module analyze different aspects of a visual stimulus. Some (e.g., in “blobs”) analyze the color of an object, while other neurons (e.g., those in “Orientation Columns”) analyze the shape or form of an object. B. Some features of cortical modules: 1. There are about 1000 of them in primary visual cortex with adjacent cortical modules analyzing inputs from adjacent locations in the visual field (forming a visuotopic “map” of the world onto primary visual cortex). 2. They are cubes (2 mm x 2 mm x 2 mm). 3. There are about 300,000 neurons in each cortical module. 4. A cortical module contains 4 ocular dominance columns (actually shaped like slabs). 5. A cortical module has 36 orientation columns (also shaped like slabs). 6. A cortical module has 16 blobs (ironically, these are shaped like columns). C. Figure 10.26 shows a diagrammed example of a cortical module. Neurons in a single cortical module all respond to visual input from the same particular location in the visual field. That is, their receptive fields all have similar locations. Understand Figure 10.26 © Hongdian Yang. This content is protected and may not be shared, uploaded, or distributed. 1. What defines an ocular dominance column is that all the neurons in it respond primarily and preferentially to input from the same eye. Neurons in layer 4C respond to input from only one eye (one hemiretina), while the neurons in the layers above and below are driven by input from both eyes (i.e., the ipsilateral temporal hemiretina and contralateral nasal hemiretina), with the input from one hemiretina having a dominant influence. 2. Cells in blobs are color sensitive but insensitive to shape. There are 16 blobs in each cortical module, 4 per ocular dominance column. Neurons in Blobs project mainly to V4, which is specialized for color perception. 3. Information about the form of an object arises mostly from cells in the orientation column, and information about the color of an object arises mostly from cell in the blobs. Somehow this is integrated. Orientation Columns A. Below is a diagram depicting a cortical module with neurons and the stimuli that drives them: © Hongdian Yang. This content is protected and may not be shared, uploaded, or distributed. The receptive field of the neurons in the cortical module approx. electrode track from fig. 10.21 Layer 4C 5 of the 36 Orientation Columns B. What drives the activity of neurons in an orientation column (i.e. what is their adequate stimulus?) 1. A spot of light or a spot of darkness will only weakly drive these neurons. 2. An elongated stimulus is what drives these neurons (a bar of light or dark of a particular orientation). © Hongdian Yang. This content is protected and may not be shared, uploaded, or distributed. 3. All the neurons in a particular orientation column respond to a line of the same particular orientation. 4. Neurons in adjacent orientation columns respond to lines with a 10 difference in rotation. Thus there is a progressive rotation of line orientation that drives these neurons as you move across the columns. If you record activity across all 36 orientation columns there will be a complete 360 rotation of the orientation of the “best stimulus.” 5. Your perception on a visual stimulus is thought to depend upon which subsets of cells within the relevant cortical modules are active. C. All this suggests that the visual world is broken down into a mosaic, with each portion of the mosaic being analyzed by a different group of cells (one cortical module analyses one piece of the mosaic and the adjacent cortical module analyses an adjacent piece of the mosaic) and what we perceive is dependent upon which subset of neurons in each cortical module is active. D. The processing of visual information in V1 is an intermediate level of processing. Now we go to a "later" stages, in inferotemporal (“ventral” stream - color and object recognition) and parietal cortex (“dorsal” stream, orientation and motion). The inferotemporal cortex (IT) A. The IT cortex has subsets of neurons that respond selectively to highly complex visual stimuli (hands, faces, food, etc.) B. How do we recognize objects as belonging to a particular category? Answer: There are specific neurons (object recognition neurons) in the IT that respond only to specific objects within a particular category (i.e. hand, faces, chairs). Researchers have recorded from neurons in IT cortex of a monkey while projecting images on a screen that the subject can see. 1. It was found that some IT neurons responded only to a particular class of images. © Hongdian Yang. This content is protected and may not be shared, uploaded, or distributed. 2. They found some neurons that responded only to faces. These neurons increased their rate of action potential production (increased their firing rate) whenever the image of a face was on the screen but not when a hand or a book or any other object was shown on the screen. 3. The possibility was tested that these neurons might actually be responding to particular portion of the face and not to the face as a whole. When these researches blanked out particular portions of the face (for example the eyes or mouth) the neurons still responded to that image with increases in firing rate. However, when the face was scrambled, the neurons no longer responded. 4. Both of these findings suggest that these are face recognition neurons, not neurons simply responding to the eye or mouth or some other subpart of the face. The above evidence also suggests that it is the spatial relationship among the subfeatures of the face that is important in inducing the neuron to respond. But what about views that dramatically change the type of image that an object casts onto the retina (for example, frontal versus profile)? Results show that a neuron that responds to a frontal view of a face responds less and less as the face is rotated to a profile view, an orientation to which it does not respond. Other neurons in IT respond to a profile view of a face but not to the frontal view. © Hongdian Yang. This content is protected and may not be shared, uploaded, or distributed. P F "Frontal" responsive neuron AP response of two IT neurons to frontal vs. profile faces "Profile" responsive neuron Frontal Intermediate Profile So, it is thought that our perception of object consistency is produced by the integration of information from neurons that respond to different views of the same object. © Hongdian Yang. This content is protected and may not be shared, uploaded, or distributed.