Lecture 07 - With Professor Darcy Kelley PDF

Preview

Summary

With Professor Darcy Kelley...

Description

   



 

  

   







Retina develops from the neural plate: minibrain Photoreceptor: 1' sensory neuron What is a receptive field? o Visual pathway Receptive field structures: RGC, LGN, V1 o Lateral geniculate nucleus - has 6 layers; switches organization right/lef o V1: has ocular dominance columns (if you go from pial -> ventricular surface, all of those cells are driven by one eye rather than the other; if you occlude one eye, the other one takes over) Visual info from photoreceptor to RGC; bipolar (two types: on/off; account for shape of receptive field of retinal ganglion cells; project to on/off retinal ganglion cells), horizontal (lateral inhibition, sharpens contrast), amacrine (starburst amacrine cells seem to be specialized [based on dendrites] for detection of motion) o RGCs tile the retina, starburst amacrine cells shingle the retina (dendritic fields overlap) Can tell where you are in the visual system using cytoarchitectonics, cortical layers Cortical organization is columnar o Chart with five neuron types, go listen to lecture again Fovea only has cones (most densely packed place in retina; highest visual acuity) Magnification factor - higher amount of neurons ____listen to lecture again__ A visuotopic map (Tootell et al, 1988) o A striped stimulus provided to eye; pulled out a part of the visual cortex; did it over and over again to change metabolism of synaptic recipient regions to upregulate production of cytochrome oxidase; Magnification factor (?) Molecules that respond to neural activity: o 2-deoxyglucose, cytochrome oxidase Look at fig 25-11 Both start out in 1' visual cortex, then they diverge (example of parallel processing) o Ventral pathway (responsible for complicated visual fields): 1' visual cortex -> inferotemporal cortex (face cell patches here respond to faces)  Inferotemporal cortex also gets input from other parts of brain How do you get a complicated receptive field? (Fig 27-10) o If you have a bunch of cells with a simpler receptive field, and they provide subthreshold input to a recipient cell, you can arrange things so that the cells only give input when visual stuff is summed o Have to have all of those receptor fields converging onto a post-synaptic cell o Many systems require convergence for specificity, but info also diverges - a single RGC provides info to a large number of circuits w/in the visual pathway that do diff things Fig 28-13 o Have complicated receptor fields in V1 (right side, horse + circus tent); have lower level stuff on right that converges/consolidates in medial temporal lobe Fig 28-5 o Object invariance - orientation of object d/n matter, can still recognize it from a different view (e.g. identifying someone from the back of their head) o Also asked if it matter whether it has color or not; found it usually doesn't matter



  











o Are there cells that identify an object as a unique object (e.g., your grandmother)? Fig 28-3 o Apperceptive agnosia - cannot see object parts as a unified whole o Associative agnosia - cannot interpret, understand, or assign meaning to objects Object blindness - cannot see an object in its entirety (can still see faces) Prosopagnosia can be acquired or developmental (birth) Fig 28-2: Proposed neural system for object recognition system o System invokes both dorsal and ventral streams; has a memory component in brain that involves hippocampus, ER, etc. Freedman and Miller o What features are necessary to recognize cat vs dog o If you visually morph cats/dogs, what makes you recognize them as cat/dog?  Record and ask whether brain response is same or different  Presumably input from TC -> lateral PFC accounts for some of it Fig 28-4 o Put electrodes in inferior temporal cortex, create peristimulus time histograms (PSTHs) o Only thing that will drive the neurons are faces (monkey/human), won't be activated by hands or scrambled faces (2) o If you can reconstruct the object from the firing pattern of the neuron, that's the HG test Fig 28-6 o Recording in ITC; there were neurons here highly selective for faces (face cell patches) o Did this with microelectrodes originally, then fMRI (yellow parts are face cell patches) o Face cell patches respond to different aspects of faces; all interconnected; if you stimulate PL, you'll see activity in other face areas Tanaka 2003 o Does it have object invariance?  Can get same activation of face cell whether you're looking from side or from front PAPER: The Code for Facial Identity in the Primate Brain o Started with a library of 200 faces, wanted to pull out info about shape of face and appearance (what is lef afer you normalize for shape)  Everyone's face shape is the same, so only thing lef is appearance o Use shape and appearance to map receptive field of face neuron o Principle component analysis: unbiased way to tell you what things in the data set matter (appearance? shape?)  Once they separated out shape, also separated out appearance (e.g., do you have bags under your eyes) o Single cells are tuned to single face axes and are blind to changes orthogonal to this axis o Using this axis model, they could predict neural firing pattern and could also read out the neural firing pattern and reconstruct the face (via PCA info) o Face patches ML/MF and AM carry complementary information about faces...