|
|
 |
Overview
My research asks how the brain makes sense of the world viewed by the eye. Insects are ideal for tackling this problem at theoretical, physiological and behavioral levels. With a visual system that accounts for as much as 30% of the lifted mass, some flying insects invest more in vision than any other animal. What happens to the abundance of information collected by such large eyes? How has the brain evolved to optimally extract the features from scenes that are most relevant to the behavior adopted?
In my lab, we employ multidisciplinary approaches in addressing these challenging questions. We concentrate on neural pathways used for detection of moving patterns and objects. Computer models for circuits that can "filter" out behaviorally relevant components of images that change in space and time are compared with physiological recordings from neurons in the insect brain. Both theory and physiological data are related to the behavior of different insects and their visual ecology. Our principal aim is to deduce the algorithms used bi insect neurons to generate specific responses to visual patterns. The insect optic lobe is a superb model system for studying mechanisms by which networks of neurons analyse visual stimuli. We adopt a wide variety of techniques drawn from biology, computer science and engineering to augment our basic neurophysiological approach to studying this system.
A large part of our current effort is directed at understanding
and modelling neurons involved in detection of moving features
and 'optical flow' patterns induced by movement by an animal through
its habitat. We are also collaborating with Industry to develop
robust models for adaptive motion detectors, based on insect vision,
for implementation in silicon hardware. Applications for this
technology include the aerospace industry, guidance systems for
miniature autonomous vehicles and for embedded collision avoidance
sensors that can be incorporated into future motor vehicles.
Recent work
Our main experimental approach is to make intracellular recordings from neurons in the optic lobes of the insect brain (i.e. measure the electrical potential inside a single neuron in the brain). Responses are then recorded to patterns presented on a high-speed (because insects are fast), computer generated display. This lets me find out what patterns produce the strongest response. By varying specific pattern attributes (such as the length of a bar or the distance between stripes in a 'grating' pattern) it is possible to say a great deal about the way in which the inputs to the neuron are wired together.
This approach has allowed us to make several interesting findings in recent years:
(1) Insects possess neurons 'tuned' for detecting specific pattern features such as oriented lines, edges or moving spots. Many of these neurons have properties remarkably similar to those described from the primary visual centres of the brain in cats or monkeys. This leads to a tempting hypothesis that similar mechanisms may underlie pattern vision in insects and mammals (see abstract). This is to some extent supported by much work on insect behaviour from other labs
(2) We have long known that some neurons in the brains
of both insects and mammals are 'tuned' to detect specific types
of moving pattern. For example, there are neurons that are excited
by motion from the front of the eye to the back, but inhibited
by back-front motion. Others prefer patterns that are rotated
in a circular fashion, or move downwards, or upwards. It is believed
that each of these neurons is optimally tuned to 'extract' different
kinds of motion that are generated as the animal moves through
its world in different ways. Thus as you move forward, the world
appears to 'flow' from front to back. As you turn, the world appears
to turn in the opposite direction. By providing information about
self-motion that can be fed-back to motor control centres, such
neurons may play a critical role in coping with unforeseen obstacles
or changes (such as sudden gusts of side-wind).
| We advanced a hypothesis that if such neurons are to optimally extract specific 'classes' of pattern motion induced by the animals own behaviour, they must be further tuned to match the speed of that motion (see abstract). Since different insects adopt different behaviour, we have used a comparative approach to test this hypothesis. By comparing properties of neurons from many insect species with different behaviour, we found a strong correlation between the optimum speed of a pattern and the lifestyle adopted by the animal. Some hovering fly species or hawkmoths need to detect very low pattern speeds while they are hovering (like hummingbirds) in front of flowers and sucking nectar with a long proboscis. They have motion detectors tuned to very low image speeds. |
The bee-fly, Bombylius |
Other insects which feed from the same flowers rarely hover
and indeed some bumblebees 'crash land' on a flower before feeding
and then rapidly darting off to the next flower. Such insects
have motion detectors tuned to very high image speeds. Other insects
adopt diverse behaviour and have correspondingly broad sensitivity
to both low and high images speeds.
(3) A common feature of motion detectors in both insects and mammals (including humans) is that they adapt to visual motion. We have begun studying motion adaptation in the context of encoding motion in natural moving scenes. Previous work on motion adaptation suggested that it served a role of optimizing motion detector responses to detect high speed pattern motion, upon prolonged exposure to such motion. extending the 'dynamic range' of a fixed motion detector mechanism. In recent years, our research has altered this view, showing that motion detectors of insects achieve snesitivity to very high image speeds without the need for an adaptive mechanism and that instead, adaptation serves a role as an automatic gain control mechanism. Our current efforts aim to refine our understandingof this gain-control, and its consequences for encoding complex motions during normal flight behavior.
Selected papers (abstracts for some papers are also available)
Harris, R.A. & O'Carroll, D.C. (2002) Afterimages in fly motion vision. Vision Research 42: 1701-1714
O'Carroll, D.C. (2001) Motion adaptation and evidence for parallel processing in the lobula plate of the bee-fly Bombylius major . In Zanker, J.M., Zeil, J., (Eds.) Motion Vision, Computational, Neural, and Ecological Constraints. Springer Verlag, Berlin Heidelberg New York 2001. pp 381-394
Dror, R.O., O'Carroll., D.C., & Laughlin, S.B. (2001) Accuracy of velocity estimation by Reichardt correlators. J. Opt. Soc. Am. A 18: 241-252
Harris, R.A., O'Carroll, D.C. & Laughlin, S.B.(2000) Contrast gain in fly motion adaptation. Neuron 28: 595-606
Tatler, B., O'Carroll, D.C., & Laughlin,S.B. (2000) Temperature, transduction and the temporal resolving power of fly photoreceptors. J. Comp. Physiol. A 186: 399-407
Dacke M., Nilsson, D.-E., Warrant, E.J., Blest, A.D., Land, M.F. & O'Carroll, D.C. (1999) Built-in polarizers form part of a compass organ in spiders. Nature (London) 401: 470-473
Harris, R.A., O'Carroll, D.C., & Laughlin, S.B.(1999), "Adaptation and the temporal delay filter of fly motion detectors", Vision Research, 39: 2603-2613.
O'Carroll, D.C. (1993), "Feature
detecting neurons in dragonflies", Nature (London)
362: pp 541-543
More information on current
research projects...
|
The University of Adelaide Insect Vision (Bugeye) Group |
 |
 (My former Lab at the Zoology Department in Cambridge) |
| Contact Me | Other insect vision labs |
last modified July 2002, © D.C.O'Carroll.