Lecture 5- Object Recognition PDF

Preview

Summary

Lecture 5, Semester 2- Object Recognition...

Description

PS20167A- COGNITIVE NEUROSCIENCE Lecture 5- Object Perception Aims and Objectives  Part 1: to provide an overview of visual processing  Part 2: to introduce the various classes of visual object recognition deficit that can occur after brain lesion  Learning objectives o To contrast the roles of the ventral and dorsal streams and illustrate the locations and functions of the visual areas o Look at some types of visual agnosia to throw light on the underlying object recognition processes Object Recognition  Object recognition is the fundamental aspect of cognition which has influence on o Memory o Decision-making o Actions  A failure to recognise something- to experience a failure of knowledge- is referred to as agnosia- when the disorder is limited to the visual modality this is visual agnosia  This has provided a window into the processes that underlie object recognition. By analysing subtypes of visual agnosia and their deficits we can draw inferences about the processes that lead to object recognition  There are four major concepts to keep in mind when thinking about object recognition: o Need to be precise when using terms like perceive or recognise- may be able to see objects but not perceive or recognise them o Although our sensory systems use a divide and conquer strategy, our perception is of unified objects- features like colour and motion are processed in distinct neural pathways however perception involves more than perceiving the features of object. o Perceptual capabilities are enormously flexible and robust  Things look the same whether people view it with both eyes or only one eye  We can also recognise things even if we see them from a different angle or if part is occluded  The percept remains stable if we stand on our head- we can attribute the change in percept to our viewing position  The product of perception is intimately interwoven with memory- we link features to form a coherent whole and that whole triggers memories The Visual System  Needed to discover the visual centre of the brain to understand how the brain represent objects  Methods to look at the visual centre of the brain- Neuropsychology o Invasive methods - single cell recordings, neural tracers, lesions, microstimulation etc. o Non-invasive methods - psychophysics, EEG, fMRI, MEG, TMS  We are now confident that the visual centre of the brain is in the occipital lobe and have discovered a lot of information about the different regions of the occipital lobe o The Primary Visual Cortex- V1- the primary cortex that receives primary visual information

o Cortical Blindness- damage to V1  Lesions to the Primary Visual Cortex results in blindness  The extent of the blindness and where it is in visual field, relates to the location and size of the cortical damage  Blindness occurs on opposite side of visual field to damage (crossing)  Patients report not being able to see anything within the blind area o Lateral Occipital Complex (LOC)  Lateral and ventral aspects of the occipital lobe  Evidence from animal and human studies have consistently shown that the LOC responds to complex object shapes (e.g., shapes, faces and 3D forms)  LOC lesions and fMRI (James, Culham, Humphreys, Milner & Goodale, 2003) result in poor object recognition- the patient lost her shape recognition system  Controls had strong activation in LOC area in line drawing but the patient did not, she had lost the ability to recognise shapes  Poor object recognition (Goodale and Milner, 1992)  Patients cannot identify the object but have good idea of it’s spatial configuration The Ventral and Dorsal Pathways  Visual processing occurs through two visual streams. o Most information goes to V1 in the occipital lobe o Output from the V1 is contained in two major fibre bundles (fasciculi) which take one of two paths  Dorsal pathway- the occipitoparietal pathway o The superior longitudinal fasciculus takes a dorsal path from the V1 to posterior regions of the parietal lobe o The dorsal stream is specialised for spatial perception- the ‘where’ pathway- helps determine where an object is by analysing spatial configurations between objects o The where system appears to be essential for determining the location of objects and also for guiding interactions with these objects o Neurons in the parietal lobe have large receptive fields that include cells representing both the fovea and the periphery o Patients with optic ataxia can recognise objects but cannot use visual information to guide action. This is associated with lesions in the parietal cortex. o Berryhill and Olson (2009)- fMRI evidence of processing of depth in parietal-occipital lesions  Both controls and those with damage asked to identify distance of objectsthose with lesions struggled to do so but could recognise the objects o Pohl (1973)- animals with parietal lobe lesions has trouble discriminating where an object was in relation to other objects.  Ventral pathway- the occipitotemporal pathway o The inferior longitudinal fasciculus follows a ventral route from the V1 to the temporal lobe o The ventral stream is specialised for object perception and recognition- the ‘what’ pathway- helps determine what we’re looking at o The receptive fields of the neurons in the ventral stream always encompass the foveastrong representation of central vision is ideal for a system devoted to object recognition



o Cells in the temporal lobe early on in processing have a preference for relatively simple features like edges whilst those further along have a preference for more complex features o The what system is essential for determining the identity of an object- if the object is familiar people will recognise it as such, if it is novel we may compare the percept to stored representations of similar objects o Patients with selective lesions of the ventral pathway may have severe problems continuously identifying objects, yet they can use the visual information to guide coordinated movement o Riddoch (1917)- Bilateral parieto-occipital lesions  To the patient a chair is flat, though he knows from experience that his visual impressions are cheating him. A stair is a flat inclined plane with no protruding steps, and yet he knows from the light and shade that he ought to see the steps. o Pohl (1973)- animals with bilateral lesions to the temporal lobe that disrupted the ventral stream had great difficulty discriminating different shapes- a what discrimination but were able to determine where objects were in relation to other objects Ventral vs. Dorsal pathways o Bell, Pasternak, Ungerleider (2014)

o Visual Recognition Deficits  Selective damage to the visual pathways gives rise to impairment of both low- and higherlevel vision (features à objects)  People with visual agnosia cannot recognise things even though visual information continues to be registered at a cortical level  There are distinct subtypes of visual agnosia which have different recognition deficits which arise at different levels of processing  Apperceptive agnosia o Can perform normally on shape discrimination tasks and can identify conventional objects o The problem becomes evident when the patient is asked to identify objects based on little stimulus information like unusual views or parts separated objects





o But able to identify conventional objects even with poor discrimination of parts o They have a problem integrating spatial information o Davidoff, J. & Warrington (1999) and Warrington & Rabin (1970)- Patients with right sided parietal lesions  Have poor recognition in perceptual matching tasks- had to determine if two stimuli were the same or different- they have a problem integrating spatial information  They also have poor recognition when presented with unusual/atypical views of objects (for example a cat viewed top down or from behind or without legs)they could not name the object shown  They could identify the objects from conventional views, confirming their problem was not due to lost visual knowledge o Their brain lacks the ability to achieve object constancy- they can recognise objects but this ability diminishes when the perceptual input is limited (for example when under shadow or when it does not include their most salient features as with atypical views) o Findings that this type of disorder are more common in patients with right-hemisphere lesions suggests this hemisphere is essential for the operations required to achieve object constancy Integrative agnosia o Patients are unable to integrate features into parts or parts of an object into a coherent whole. o They are able to perform shape-matching tasks and are able to match images of objects seen from unusual views. o Rather than perceiving objects at a glance they have to perceive salient features and parts and try to put them together step by step to identify the whole object. o When objects overlap the observer may struggle to identify parts and correctly assign them to objects o Behrmann, Moscovitch, & Winocur, 1994  Man asked to replicate a picture which consisted of two diamonds and a circle that were touching- he failed to recognise that the lines linked into recognisable wholes- the order he drew the lines showed he did not recognise the separate shapes o Copying without normal object recognition, along with (initially) spared stored visual memory Associative agnosia o Associative agnosia – can perceive objects but has a problem with the association between the visual object recognition and its semantic links o McCarthy & Warrington (1986)  Able to colour in the different objects of complex drawings, BUT cannot identify the different objects coloured  If the objects were named he could point to them effortlessly

 



Recognition problems reflected inability to access semantic knowledge from the visual modality- they derive normal visual representations but cannot use this information to recognise things o Category specificity in Agnosia- sometimes there can be selective loss within the knowledge system such that certain categories of objects (e.g. living ones vs. inanimate ones) are not recognised but others are.  Patients with category specific deficits support the idea that lesions and brain damage do not completely destroy the connections to semantic knowledge but may destroy areas devoted to processing similar types of information  Common issue where agnosia more likely to affect living than inanimate objects may be due to manufactured objects activating different forms of representation (likely to have held and used a telephone, not a tiger)- these extra representations of interactions with objects may be sufficient to allow someone to recognise non-living objects  Category specific deficits may reflect the organisation of semantic memory knowledge Prosopagnosia o Prosopagnosia is a failure to recognise faces o Pallis’s (1955) patient put it:  “I can see the nose, and mouth quite clearly but they just don’t add up. They all seem chalked in, like on a blackboard. I have to tell by the clothes or voice whether it is a man or a woman”  Couldn’t even recognise himself- a face he would see in the mirror every day  In other ways the patient has excellent memory, recognised common objects without hesitation and could read and recognise all line drawings o Often results from lesions of the fusiform gyrus of the temporal lobe- Riddoch, Johnston, Bracewell, Boutsen & Humphreys (2008)  McCarthy et al (1997)- participants showed stronger BOLD response in the fusiform gyrus when viewing faces than any other stimuli  Led to it being referred to as the fusiform face area o Superior temporal sulcus also implicated in more dynamic features of face processingit is responsive to facial emotions and is also more active during lip reading or monitoring eye gaze (Haxby et al, 2000)  Jiang & He, 2006- fusiform face area was activated in response to all faces but the superior temporal sulcus responded only to emotive faces. o Thatcher effect- we don’t tend to notice that when presented upside down, Thatcher’s eyes and mouth have been inverted- suggests we view faces primarily holistically

Mind Reading  Decoding models can be used to predict the stimulus someone is viewing based on a physiological response such as BOLD activity across a set of voxels



Mind reading may one day be used to communicate with people who are currently unable to speak...