The representation and processing of contour and surface in the human brain

The visual brain has been interrogated with functional magnetic resonance imaging (fMRI) over the last twenty years, providing significant insight into its organisation. Earlier processing occurs in posterior, 'lower-level' visual brain areas, which contain explicit visual field maps. Such areas are 'mapped' by using visual stimuli that systematically move through the visual field, whilst recording the brain responses. Improvements in techniques have allowed an increasing extent of the visual brain to be mapped in this manner. An alternative method to chart the visual brain's organisation is to test for stimulus selectivity, typically by contrasting the brain's responses to two or more discrete stimulus categories. This approach has identified 'higher-level' category selective areas, for example a shape-selective region referred to as the lateral occipital complex.

It was first thought that category selective regions had little or no explicit mapping of the visual field. However, as more visual field maps were discovered some were found to overlap with category selective regions, implying that the two organizations are not mutually exclusive. We will refer to these overlapping regions as 'mid-level' visual field maps. The fact that the two organizing principles intersect gives us an opportunity to understand how the visual brain transitions from low-to-high level processing. Such transitional points should yield valuable insights into the representation of 'abstract' information in higher-level areas.

Despite their potential importance, little research on these mid-level regions exists. We aim to address this by focusing on understanding the roles 'mid-level' visual areas play. We have published evidence that mid-level areas can represent abstract information, but the nature of the information represented at this level and the nature of the representations themselves remain unknown. Our overarching aim, therefore, is to find out what visual information is processed and how it represented at this mid-level of the visual system.

The mid-level visual areas we will focus on cluster in lateral occipital cortex. At the same time we will also assess visual field maps that lie in dorsal and ventral occipital cortex, to verify that any effects we measure are dissociable. Reassuringly, using this strategy we have already found that certain abstract representations are found in lateral, but not ventral or dorsal mid-level areas. We anticipate therefore that our work will again exhibit dissociable effects in mid-level areas.

Our work implies that lateral occipital areas process aspects of contour curvature; we aim to further refine our understanding of these representations. We already have pilot data on some aspects of contour processing and believe therefore that much of the proposed work has a high chance of success. At the same time however we want to explore the processing and representation of 3D surface curvature, to determine whether such properties have shared or unique neural underpinnings. This work is higher risk, but its return could be large as it represents the first opportunity to bring together the study of contour and surface and learn how their relative contributions to our perception of objects is achieved in the human brain.

We will combine perceptual measurements and fMRI to reveal how the human brain processes and represents contour and surface curvature. The perceptual experiments will reveal how our stimulus manipulations are perceived; determining which stimuli will be presented during neuroimaging experiments. Both approaches will leverage our expertise in stimulus control, which is grounded in computational methods that can generalize from contour to surface. Further, we have demonstrated that we can characterize our stimuli with perceptual and physical metrics that explain patterns of response in mid-level visual areas and this approach will be extended in the proposed work.

The visual cortex is organized into multiple visual field maps that cluster in anatomical zones. The early visual maps V1-3 cluster in the posterior occipital cortex, whereas mid-level areas cluster in lateral (LO-1 & LO-2) and ventral occipital (V4, VO-1 & VO-2) cortex. There is homology between human and macaque monkey in the arrangement of early visual areas (V1-3), but the homology breaks down in mid-level areas and it has been argued that areas LO-1 and LO-2 are unique to human. Unraveling what these two areas do in the context of visual spatial processing is important therefore as there is no animal model to refer to. Similarly, much of the visual spatial processing undertaken in macaque V4 may not, as has been previously assumed, occur in human V4. There is an imperative therefore to explore the processing and representations of visual spatial information in both ventral and lateral occipital areas to uncover what respective roles they play in object recognition.

We will focus on the roles of lateral occipital areas and compare the processing and representation of visual spatial properties within them with other mid-level areas in ventral occipital regions. Our work to date has shown how mid-level areas serve an important role in the transition from two organizational principles of visual field mapping to category selectivity. We wish to further this work by assessing the roles mid-level areas play in the processing and representation of contour and surface curvature.

We have chosen to investigate curvature of contours and surface because they are important cues to object identity. We can also mathematically define such stimuli precisely and capture their perceptual and physical characteristics to use as predictors for representational similarity analysis of mid-level visual areas. This offers us an advantage of teasing apart what unique contributions the neural processing and representations in mid-level areas make in terms of our perception.

The work in this proposal is basic science; it aims to understand the neural underpinnings of human visual perception. The work will specifically advance our knowledge of how neural activity in mid-level areas of the brain contributes to the perception of shape and objects. The outcomes may be of potential interest to a variety of different user groups:

- computational neuroscientists interested in biologically inspired models of shape and object processing,

- engineers interested in artificial vision,

- clinicians who deal with patients with visual deficits caused by brain damage

- the general public who have a general interest in how the human brain works.

Commercial and Clinical Stakeholders: The immediate commercially exploitable outcomes of this research may not be obvious, but we will engage with commercial partners (some of which we have established links with) in order to identify potential avenues for development. In order to reach these groups and ensure that any outcomes of this research are fully exploited the pathways we intend to use run a workshop, which will be held at the end of year two of the proposed funded period. The aim of these symposia will be to engage technology industries and researchers in the areas of computational vision and robotics in order identify the commercial and scientifically exploitable aspects of this research. In a separate effort to engage we will present work at the annual meetings we hold on vision (the Yorkshire Vision Network), which bring together clinical neurologists and neuropsychologists in order to explore fully and exploit the possible clinical benefits of the proposed research.

Interactive Public Engagement Activity: The human brain is a subject of great interest to the public (at all ages). Our research provides an opportunity to inform a wider non-scientific audience about the neural processing that underpins visual perception in humans. This will be achieved via public lectures and schools visits and will be a continuation of the applicants' established working practices. Specific to this project we also intend to augment our public facing website which will describe our experiments and provide resources about brain function and perception of contour and surface.

In order to facilitate impact outside the immediate research area of the study the applicants will work in conjunction with the University of York's Research and Enterprise Office. Their expertise will be used to identify any possibilities in this area and develop any possible commercial potential. Alongside clinical, academic and commercial end users of this research members of these offices will be asked to sit on the workshop described above, allowing them to provide guidance on impact activities. These impact pathways will operate in parallel with the standard methods of engagement with our academic colleagues through publication in peer reviewed journals, depositing data and models in public repositories and by presentation of our work at meetings and in invited talks.