How Does Human Ventral Temporal Cortex Contribute to Object Perception
New Aims rev 5/2/2008
Specific Aims Object recognition is one of the most important and challenging aspects of visual perception: human are able to rapidly and reliably identify a wide range of objects, both novel and familiar. A variety of experimental approaches suggest that the human ventral temporal cortex (VTC) is critical to object recognition, but little is known about how neuronal activity in this region contributes to perception of complex forms. Most studies of VTC use blood-oxygen level dependent functional magnetic resonance imaging (BOLD fMRI). BOLD fMRI is a powerful technique because it allows a global survey of activity across all parts of VTC. However, it suffers from a critical limitation: because it measures a hemodynamic, not a neural signal, it is several steps removed from the neuronal activity that we are actually interested in. The proposed experiments capitalize on a unique opportunity to directly record electrical activity in small populations of ventral temporal neurons using patients implanted with subdural surface electrodes. By combining fMRI and intracranial recording in the same human subjects, we can obtain the best of both worlds. An accurate assessment of precisely what visual area is being recorded from; and a direct measurement of neuronal activity in these areas. This will allow us to make significant advances in our ounderstanding of human VTC. first characterize ventral temporal cortical visual areas with fMRI in individual human subjects, and then use electrodes subsequently implanted in the same subjects to investigate neuronal activity in identified visual areas in VTC in response to specific visual stimuli and in relation to behavior performance on object recognition tasks.
Specific Aim 1: To what degree are small populations of neurons in VTC selective for specific visual objects? Human fMRI studies show that discrete regions in VTC respond selectively to specific categories of objects such as faces or places. However, the functional organization of VTC that underlies these preferences and the detailed nature of the preferred stimuli and the category specificity throughout VTC has not been well characterized. We will record neuronal responses in object selective VTC to a wide range of stimuli to identify preferred stimuli for specific regions of VTC, and determine how broadly tuned are small groups of VTC neurons to stimulus category, and to specific exemplars within categories. We will then measure the effect that repeated presentations of a preferred stimulus has on stimulus-induced response properties in order to determine if the recorded responses primarily reflect the sampling of a localized collection of highly selectively tuned neurons (which would be expected to show repetition suppression), or that the individual neurons that collectively generate the recorded LFP are broadly tuned across categories or exemplars. Hypothesis: Small populations of human VTC neurons are broadly tuned to respond to a range of specific classes of objects, consistent with a distributed coding model.
Specific Aim 2: How large are the spatial RFs in VTC, and how is spatial tuning related to stimulus category preference? Visual areas in late stages of the ventral stream of visual processing, such as object selective human VTC, are thought to be highly tuned for object recognition and in turn highly tolerant of variations in stimulus position. It has further been hypothesized that spatial characteristics of neurons in VTC have relate to stimulus preferences, with face responsive neurons (FFA) having foveal RFs and place (PPA) having peripheral RFs. While fMRI does not readily allow direct measure of spatial RF, our recording techniques in human subjects are well-suited to these studies. We will record stimulus-evoked responses from implanted electrodes in human subjects to quantitatively measure spatial RF properties in VTC. Furthermore, we will determine the relationship between stimulus category (face vs. place) and spatial RF size. Hypothesis: VTC is broadly tolerant, but not invariant, to variations in stimulus position irrespective of object category.
Specific Aim 3: How is behavioral variability during an object recognition task related to local neuronal response variability? One promising approach in studies of animal subjects has been the exploration of how the strength of the physiological responses of sensory neurons correlates with the behavioral responses on a trial-by-trial basis. We will apply this approach to recordings of field potentials form cortical surface electrodes in human VTC to determine if the magnitude of responses correlates with behavioral reposes, whether this correlation is specific for stimulus conditions for which the specific VTC recording site is likely to be most important. Such a correlation would provide strong evidence that VTC plays a critical role in object perception. Hypothesis: The strength of responses to preferred (but not unpreferred) objects will correlate significantly with behavioral object recognition.
Specific Aim 4: How does direct activation of identified visual areas in VTC influence perception? 4.1 We know that stimulation of early visual cortex produces simple phosphenes, which correspond with the relatively simple RF properties of early visual cortex. RFs in VTC are much more complex (selective for faces, houses, etc.). Does stimulation of neurons with more complex RFs ever produce complex percepts that match the selectivity? Does selectivity match perception in VTC? 4. 2: If stimulation does not produce of a later area does not produce a percept, can it bias perception of the objects for which that electrode is selective While the demonstration of the selectivity of VTC neurons for specific objects, and the demonstration of significant correlations between VTC physiological responses and object recognition on a trial-by-trial basis both offer strong evidence for VTC playing a crucial role in human object perception, these findings would fall short of demonstrating that activity of VTC in itself is capable of generating or influencing the perception of complex forms such as faces. To test this hypothesis, we will electrically stimulate identified object selective regions of VTC while subjects perform an object recognition task and measure the influence of indirect activation of VTC on object perception. Hypothesis: Electrical activation of VTC cortex will interfere the subject’s ability to perceive the specific objects that the stimulated population of neurons is selective for, providing the strongest evidence to date of the critical role this area plays in object perception.
Specific Aim 1: Identify the preferred stimulus and stimulus categories of small populations of neurons in ventral temporal cortex? Our hypothesis is that small populations of neurons will prefer specific classes of stimuli. Questions: how broadly-tuned are the neurons? How spatially-selective are the neurons?
Specific Aim 2: How does the activity in small populations of neurons correlate with behavior? Our hypothesis (see DY K-award aim 3) is that heightened activity will make subjects more likely to percept an object.
Specific Aim 3: Identify the effect of artificially-induced activity in ventral temporal cortex on object perception. Our hypothesis is that interfering with the activity in ventral temporal cortex will interfere with the ability of the subject to perceive the objects that the small population of neurons is selective for.
Human Background data: Selective BOLD responses to faces or objects in FFA. PPA, LOC Lesions in bilateral ventral temporal cortex have been associated with agnosia for objects
Monkey Background data: Numerous studies of IT show selective responses to higher order objects Affraz showed that indirect activation of IT alters perception of objects.
Preliminary data: Using fMRI, we can map retinotopic cortex and LOC/VTC in patients LFPs are a good measure of neuronal activity--e.g. can record small RFs in V1 Object selective responses exist in VTC VTC responses are moderately tolerant of changes in stim size and position Stimulation of object selective VTC does not result in perceptual event (except in rare instances)
- 1: HUMAN subjects
- 2: Combine fMRI, recording and stimulation in the same brain
- 3: Stimulation advantage: use stimulation to prove necessity
- 4: Electrical recording advantages (vs. fMRI):Temporal measures are valuable
vs. EEG/MRG: precise spatial localization
Aims (revised 2/14/2008) The Role of Ventral Temporal Cortex in Visual Perception Studied with LFPs, fMRI and electrical stimulation
1) Object Responses in VTC. i) Object Invariance in VTC (what is DiCarlo's new way of phrasing invariance?) Test the hypothesis that VTC will respond to the same stimuli presented in different places ii) Selectivity in VTC Do areas that show category selectivity in fMRI also show category selectivity in LFP? If so, how tightly tuned are the responses (1 face/many faces/etc.) [[optional iii) Adaptation in VTC does repeated presentation of the preferred stimuli adapt the response or sharpen it? what about non-preferred? ]] iv) Bias perception towards/away from preferred stimulus.
2) Color Responses in VTC - areas that prefer color to B/W stimuli in fMRI, also like color with LFPs? - in LFP, prefer a certain color, or respond to all colors? - produce color percepts when stimulated? - if not, induce perception of color in gray stimuli? - does selectivity match perception
1. Nature of Visual Responses in VTC a) object selectivity in small groups of neurons in human VTC measured with LFP and fMRI Hypothesis: areas that prefer a specific stimulus category in fMRI should also prefer that category with LFPs Selectivity should be more narrowly tuned with LFPs
b) Spatial receptive fields for VTC Hypothesis: RFs should be large and contralateral in VTC; may be hot spots
2. Does stimulation of VTC object selective areas recordings evoke a visual percept? Hypothesis: Except in rare cases, will NOT be percepts because objects are represented in a sparse, distributed way.
3. Does stimulation of an object selective area of VTC result in a change in the perception of an object? Hypothesis: Yes, stimulation of areas that prefer a stimulus will ONLY affect perception of that stimulus not others.
(Aim 2 is relevant bc may need to be sure that stim is eccentric to see effect of unilateral stimulation).
Protocol fMRI: find category-object areas
LFP: map selectivity, find pref stimuli
Stim: Stimulate and see if we can bias perception
This lets us address the nature of object representation in human ventral temporal cortex. We can assess the necessity of a given area for perception of the preferred or non-preferred stimulus by disrupting or artificially activating it.
Goals for the future: Need to use smaller and more closely spaced electrodes, may exceed our methods
Collectively, these experiments will capitalize on a unique opportunity to obtain high-quality neurophysiological recordings directly from the human brain to provide new insights about the representation of objects in visual cerebral cortex and how those representations relate to behavioral performance.