On the neuronal basis of figure-ground discrimination by relative motion in the visual system of the fly

A fly can discriminate an object (figure) from its background on the basis of motion information alone. This information processing task has been analysed, so far, mainly in behavioural studies but also in electrophysiological experiments (Reichardt et al., 1983). The present study represents a further attempt to bridge the gap between the behavioural and the neuronal level. It is based on behavioural and electrophysiological experiments as well as on computer simulations. The characteristic properties of figureground discrimination behaviour impose specific constraints on the spatial integration properties of the output cells of the underlying neuronal network, the heterolateral interactions in their input circuitry, as well as on the range of variability of their response. These constraints are derived partly from previous behavioural studies (Reichardt et al., 1983), partly, however, from behavioural response characteristics which have not been addressed explicitly so far. They are interpreted in terms of one of the alternative model circuits shown by Reichardt et al. (1983) to be sufficient to account for figure-ground discrimination. It will be demonstrated, however, that this can be done equally well by means of a further alternative model circuit. These constraints are used in the electrophysiological analysis for establishing visual interneurones as output elements of the neuronal network underlying figure-ground discrimination.In the behavioural experiments on figure-ground discrimination as well as on the optomotor course control the yaw torque generated by the tethered flying fly under visual stimulation was used as a measure for the strength and time course of the reaction. Therefore, it has initially been proposed that the three Horizontal Cells, which are regarded as the output elements of the neuronal network underlying the optomotor reaction (e.g. Hausen, 1981), might also control yaw torque generation in figure-ground discrimination (Reichardt et al., 1983). New behavioural data show, however, that the Horizontal Cells do not meet all the constraints imposed on the presumed output cells of the figure-ground discrimination network: (1) The Horizontal Cells are not sensitive enough to motion of small objects. (2) The heterolateral interactions within their input circuitry are not in accordance with the behavioural data (see also Reichardt et al., 1983). (3) The variability found in the time course of certain components of the yaw torque response to relative motion of figure and ground cannot be explained by their response characteristics. Hence, the Horizontal Cells cannot account for figure-ground discrimination on their own and additional output cells of the optic lobes with different functional properties are required to accomplish this task.     

Correlation versus gradient type motion detectors: the pros and cons

Visual motion contains a wealth of information about self-motion as well as the three-dimensional structure of the environment. Therefore, it is of utmost importance for any organism with eyes. However, visual motion information is not explicitly represented at the photoreceptor level, but rather has to be computed by the nervous system from the changing retinal images as one of the first processing steps. Two prominent models have been proposed to account for this neural computation: the Reichardt detector and the gradient detector. While the Reichardt detector correlates the luminance levels derived from two adjacent image points, the gradient detector provides an estimate of the local retinal image velocity by dividing the spatial and the temporal luminance gradient. As a consequence of their different internal processing structure, both the models differ in a number of functional aspects such as their dependence on the spatial-pattern structure as well as their sensitivity to photon noise. These different properties lead to the proposal that an ideal motion detector should be of Reichardt type at low luminance levels, but of gradient type at high luminance levels. However, experiments on the fly visual systems provided unambiguous evidence in favour of the Reichardt detector under all luminance conditions. Does this mean that the fly nervous system uses suboptimal computations, or is there a functional aspect missing in the optimality criterion? In the following, I will argue in favour of the latter, showing that Reichardt detectors have an automatic gain control allowing them to dynamically adjust their input–output relationships to the statistical range of velocities presented, while gradient detectors do not have this property. As a consequence, Reichardt detectors, but not gradient detectors, always provide a maximum amount of information about stimulus velocity over a large range of velocities. This important property might explain why Reichardt type of computations have been demonstrated to underlie the extraction of motion information in the fly visual system under all luminance levels.     

Dynamic response properties of movement detectors: Theoretical analysis and electrophysiological investigation in the visual system of the fly

Dynamic aspects of the computation of visual motion information are analysed both theoretically and experimentally. The theoretical analysis is based on the type of movement detector which has been proposed to be realized in the visual system of insects (e.g. Hassenstein and Reichardt 1956; Reichardt 1957, 1961; Buchner 1984), but also of man (e.g. van Doorn and Koenderink 1982a, b; van Santen and Sperling 1984; Wilson 1985). The output of both a single movement detector and a one-dimensional array of detectors is formulated mathematically as a function of time. The resulting movement detector theory can be applied to a much wider range of moving stimuli than has been possible on the basis of previous formulations of the detector output. These stimuli comprise one-dimensional smooth detector input functions, i.e. functions which can be expanded into a time-dependent convergent Taylor series for any value of the spatial coordinate.The movement detector response can be represented by a power series. Each term of this series consists of one exclusively time-dependent component and of another component that depends, in addition, on the properties of the pattern. Even the exclusively time-dependent components of the movement detector output are not solely determined by the stimulus velocity. They rather depend in a non-linear way on the weighted sum of the instantaneous velocity and all its higher order time derivatives. The latter point represents another reason — not discussed so far in the literature — that movement detectors of the type analysed here do not represent pure velocity sensors.The significance of this movement detector theory is established for the visual system of the fly. This is done by comparing the spatially integrated movement detector response with the functional properties of the directionally-selective motion-sensitive. Horizontal Cells of the third visual ganglion of the fly's brain.These integrate local motion information over large parts of the visual field. The time course of the spatially integrated movement detector response is about proportional to the velocity of the stimulus pattern only as long as the pattern velocity and its time derivatives are sufficiently small. For large velocities and velocity changes of the stimulus pattern characteristic deviations of the response profiles from being proportional to pattern velocity are predicted on the basis of the detector theory developed here. These deviations are clearly reflected in the response of the wide-field Horizontal Cells, thus, providing very specific evidence that the movement detector theory developed here can be applied to motion detection in the fly. The characteristic dynamic features of the theoretically predicted and the experimentally determined cellular responses are exploited to estimate the time constant of the movement detector filter.     

Columnar cells necessary for motion responses of wide-field visual interneurons in <Emphasis Type="Italic">Drosophila</Emphasis>

Wide-field motion-sensitive neurons in the lobula plate (lobula plate tangential cells, LPTCs) of the fly have been studied for decades. However, it has never been conclusively shown which cells constitute their major presynaptic elements. LPTCs are supposed to be rendered directionally selective by integrating excitatory as well as inhibitory input from many local motion detectors. Based on their stratification in the different layers of the lobula plate, the columnar cells T4 and T5 are likely candidates to provide some of this input. To study their role in motion detection, we performed whole-cell recordings from LPTCs in Drosophila with T4 and T5 cells blocked using two different genetically encoded tools. In these flies, motion responses were abolished, while flicker responses largely remained. We thus demonstrate that T4 and T5 cells indeed represent those columnar cells that provide directionally selective motion information to LPTCs. Contrary to previous assumptions, flicker responses seem to be largely mediated by a third, independent pathway. This work thus represents a further step towards elucidating the complete motion detection circuitry of the fly.     

The Intrinsic Electrophysiological Characteristics of Fly Lobula Plate Tangential Cells: III. Visual Response Properties

In this last paper in a series (Borst and Haag, 1996; Haag et al., 1997) about the lobula plate tangential cells of the fly visual system (CH, HS, and VS cells), the visual response properties were examined using intracellular recordings and computer simulations. In response to visual motion stimuli, all cells responded mainly by a graded shift of their axonal membrane potential. While ipsilateral motion resulted in a graded membrane potential shift, contralateral motion led to distinct EPSPs. For HS cells, simultaneous extracellular recorded action potentials of a spiking interneuron, presumably the H2 cell, corresponded to the EPSPs in the HS cell in a one-to-one fashion. When HS cells were hyperpolarized during ipsilateral motion, they mainly produced action potentials, but when they were hyperpolarized during contralateral motion only a slight increase of EPSP amplitude, could be observed. Intracellular application of the sodium channel blocker QX 314 abolished action potentials of HS cells while having little effect on the graded membrane response to ipsilateral motion. HS and CH cells were also studied with respect to their spatial integration properties. For both cell types, their graded membrane response was found to increase less than linearly with the size of the ipsilateral motion pattern. However, while for HS cells various amounts of hyperpolarizing current injected during motion stimulation led to different saturation levels, this was not the case for CH cells. In response to a sinusoidal velocity modulation, CH cells followed pattern motion only up to 10 Hz modulation frequency, but HS cells still revealed significant membrane depolarizations up to about 40 Hz.In the computer simulations, the compartmental models of tangential cells, as derived in the previous papers, were linked to an array of local motion detectors. The model cells revealed the same basic response features as their natural counterparts. They showed a response saturation as a function of stimulus size. In CH-models, however, the saturation was less pronounced than in real CH-cells, indicating spatially nonuniform membrane resistances with higher values in the dendrite. As in the experiments, HS models responded to high-frequency velocity modulation with a higher amplitude than did CH models.     

Temporal modulation of luminance adapts time constant of fly movement detectors

The time constant of movement detectors in the fly visual system has been proposed to adapt in response to moving stimuli (de Ruyter van Steveninck et al. 1986). The objective of the present study is to analyse, whether this adaptation can be induced as well, if the luminance of a stationary uniform field is modulated in time. The experiments were done on motion-sensitive wide-field neurones of the lobula plate, the posterior part of the third visual ganglion of the blowfly, calliphora erythrocephala. These cells are assumed to receive input from large retinotopic arrays of movement detectors. In order to demonstrate that our results concern the properties of the movement detectors rather than those of a particular wide-field cell we recorded from two different types of them, the H1- and the HSE-cell. Both cell types respond to a brief movement stimulus in their preferred direction with a transient excitation. This response decays about exponentially. The time constant of this decay reflects, in a first approximation, the time constant of the presynaptic movement detectors. It was determined after prestimulation of the cell by the following stimuli: (a) periodic stationary grating; (b) uniform field, the intensity of which was modulated sinusoidally in time (flicker stimulation); (c) periodic grating moving front-to-back; (d) periodic grating moving back-to-front. The decay of the response is significantly faster not only after movement but also after flicker stimulation as compared with pre-stimulation with a stationary stimulus. This is interpreted as an adaptation of the movement detector's time constant. The finding that flicker stimulation also leads to an adaptation shows that movement is not necessary for this process. Instead the adaptation of the time constant appears to be governed mainly by the temporal modulation (i.e., contrast frequency) of the signal in each visual channel.     

Response properties of motion-sensitive visual interneurons in the lobula plate of Drosophila melanogaster

The crystalline-like structure of the optic lobes of the fruit fly Drosophila melanogaster has made them a model system for the study of neuronal cell-fate determination, axonal path finding, and target selection. For functional studies, however, the small size of the constituting visual interneurons has so far presented a formidable barrier. We have overcome this problem by establishing in vivo whole-cell recordings from genetically targeted visual interneurons of Drosophila. Here, we describe the response properties of six motion-sensitive large-field neurons in the lobula plate that form a network consisting of individually identifiable, directionally selective cells most sensitive to vertical image motion (VS cells). Individual VS cell responses to visual motion stimuli exhibit all the characteristics that are indicative of presynaptic input from elementary motion detectors of the correlation type. Different VS cells possess distinct receptive fields that are arranged sequentially along the eye's azimuth, corresponding to their characteristic cellular morphology and position within the retinotopically organized lobula plate. In addition, lateral connections between individual VS cells cause strongly overlapping receptive fields that are wider than expected from their dendritic input. Our results suggest that motion vision in different dipteran fly species is accomplished in similar circuitries and according to common algorithmic rules. The underlying neural mechanisms of population coding within the VS cell network and of elementary motion detection, respectively, can now be analyzed by the combination of electrophysiology and genetic intervention in Drosophila.     

An analog VLSI implementation of a visual interneuron: enhanced sensory processing through biophysical modeling

Flies are capable of rapid, coordinated flight through unstructured environments. This flight is guided by visual motion information that is extracted from photoreceptors in a robust manner. One feature of the fly's visual processing that adds to this robustness is the saturation of wide-field motion-sensitive neuron responses with increasing pattern size. This makes the cell's responses less dependent on the sparseness of the optical flow field while retaining motion information. By implementing a compartmental neuronal model in silicon, we add this "gain control" to an existing analog VLSI model of fly vision. This results in enhanced performance in a compact, low-power CMOS motion sensor. Our silicon system also demonstrates that modern, biophysically-detailed models of neural sensory processing systems can be instantiated in VLSI hardware.     

Propagation of photon noise and information transfer in visual motion detection

The extraction of the direction of motion from the time varying retinal images is one of the most basic tasks any visual system is confronted with. However, retinal images are severely corrupted by photon noise, in particular at low light levels, thus limiting the performance of motion detection mechanisms of what sort so ever. Here, we study how photon noise propagates through an array of Reichardt-type motion detectors that are commonly believed to underlie fly motion vision. We provide closed-form analytical expressions of the signal and noise spectra at the output of such a motion detector array. We find that Reichardt detectors reveal favorable noise suppression in the frequency range where most of the signal power resides. Most notably, due to inherent adaptive properties, the transmitted information about stimulus velocity remains nearly constant over a large range of velocity entropies. Action editor: Matthew Wiener     

Internal structure of the fly elementary motion detector

Recent experiments have shown that motion detection in Drosophila starts with splitting the visual input into two parallel channels encoding brightness increments (ON) or decrements (OFF). This suggests the existence of either two (ON-ON, OFF-OFF) or four (for all pairwise interactions) separate motion detectors. To decide between these possibilities, we stimulated flies using sequences of ON and OFF brightness pulses while recording from motion-sensitive tangential cells. We found direction-selective responses to sequences of same sign (ON-ON, OFF-OFF), but not of opposite sign (ON-OFF, OFF-ON), refuting the existence of four separate detectors. Based on further measurements, we propose a model that reproduces a variety of additional experimental data sets, including ones that were previously interpreted as support for four separate detectors. Our experiments and the derived model mark an important step in guiding further dissection of the fly motion detection circuit.     

Properties of individual movement detectors as derived from behavioural experiments on the visual system of the fly

The performance of the fly's movement detection system is analysed using the visually induced yaw torque generated during tethered flight as a behavioural indicator. In earlier studies usually large parts of the visual field were exposed to the movement stimuli; the fly's response, therefore, represented the spatially pooled output signals of a large number of local movement detectors. Here we examined the responses of individual movement detectors. The stimulus pattern was presented to the fly via small vertical slits, thus, nearly avoiding spatial integration of local movement information along the horizontal axis of the eye. The stimulus consisted of a vertically oriented sine-wave grating which was moved with a constant velocity either clockwise or counterclockwise. In agreement with the theory of movement detectors of the correlation type, the time-course of the detector signal is modulated with the spatial phase of the stimulus pattern. It can even assume negative values for some time during the response cycle and thus signal the wrong direction of motion. By spatially integrating the response over sufficiently large arrays of movement detectors these response modulations disappear. Finally, one obtains a signal of the movement detection system which is constant while the pattern moves in one direction and only changes its sign when the pattern reverses its direction of motion. Spatial integration thus represents a simple means to obtain a meaningful representations of motion information.     

Afterimage-like effects in the motion-sensitive neuron H1

A powerful effect resembling an afterimage is demonstrated on the pathway to the motion-sensitive neuron H1. This effect is independent of the locally generated gain control described in an earlier paper (Maddess & Laughlin 1985, Proc. R. Soc. Lond. B 225, 251). The afterimage, produced across the eye by a stationary pattern, causes the sensitivity to movement to be different according to the local stimulus history, and the effects of low-contrast (0.1) patterns, presented for as little as a few hundred milliseconds, remain for up to 2 s. Moving patterns interact with the afterimage to modulate the spike rate of H1. The afterimage increases with contrast but saturates at contrasts above 0.5. Low spatial frequencies generate afterimages less effectively than moderate ones; this result indicates that the afterimage process could lie at, or after, lateral inhibition between tonic units. This is supported by the fact that the altered sensitivity profiles generated by single bright and dark vertical bars initially resemble Mach bands. However, this character alters as the afterimage decays, and the depression of H1's response to moving bright stimuli, produced by the afterimage of a dark bar, continues to grow for up to 1 s after the adapting bar is removed. A short-lived (0.5 s) reduction of H1's directional selectivity accompanies strong afterimage formation. All these factors, especially the saturation at low contrasts and the spatial frequency tuning, rule out light adaptation by photoreceptors as the afterimage source. Luminances used were also low enough to exclude influence by the pupil mechanism. Lastly, responses to patterns that are occasionally jumped by large or small distances are broadened by stimuli that produce an afterimage. Responses to small displacements have previously been described as 'velocity impulse responses' (Srinivasan 1983, Vision Res. 23, 659; Zaagman et al. 1983, IEEE Trans. SMC 13, 900) and so the response broadening (stimulus blurring) can be taken as a reduction of the fly's temporal resolution of moving objects. Previously reported work shows that afterimages seen in humans and the effect reported here act over the same range of temporal frequencies rather than retinal drift speeds. This may suggest an important role for afterimage-like effects in the processing of the low temporal frequency components of moving images. Certainly, the fly's afterimage system reduces the visibility of moving objects within patches of an image that, have on average, contained slowly varying motion signals.(ABSTRACT TRUNCATED AT 400 WORDS)     

Insect-inspired high-speed motion vision system for robot control

The mechanism for motion detection in a fly’s vision system, known as the Reichardt correlator, suffers from a main shortcoming as a velocity estimator: low accuracy. To enable accurate velocity estimation, responses of the Reichardt correlator to image sequences are analyzed in this paper. An elaborated model with additional preprocessing modules is proposed. The relative error of velocity estimation is significantly reduced by establishing a real-time response-velocity lookup table based on the power spectrum analysis of the input signal. By exploiting the improved velocity estimation accuracy and the simple structure of the Reichardt correlator, a high-speed vision system of 1?kHz is designed and applied for robot yaw-angle control in real-time experiments. The experimental results demonstrate the potential and feasibility of applying insect-inspired motion detection to robot control.     

Adaptation accentuates responses of fly motion-sensitive visual neurons to sudden stimulus changes

Adaptation in sensory and neuronal systems usually leads to reduced responses to persistent or frequently presented stimuli. In contrast to simple fatigue, adapted neurons often retain their ability to encode changes in stimulus intensity and to respond when novel stimuli appear. We investigated how the level of adaptation of a fly visual motion-sensitive neuron affects its responses to discontinuities in the stimulus, i.e. sudden brief changes in one of the stimulus parameters (velocity, contrast, grating orientation and spatial frequency). Although the neuron''s overall response decreased gradually during ongoing motion stimulation, the response transients elicited by stimulus discontinuities were preserved or even enhanced with adaptation. Moreover, the enhanced sensitivity to velocity changes by adaptation was not restricted to a certain velocity range, but was present regardless of whether the neuron was adapted to a baseline velocity below or above its steady-state velocity optimum. Our results suggest that motion adaptation helps motion-sensitive neurons to preserve their sensitivity to novel stimuli even in the presence of strong tonic stimulation, for example during self-motion.     

Effects of Mean Firing on Neural Information Rate

We investigated the effect of mean firing on the information rate of a spiking motion-sensitive neuron in the fly (H1-cell). In the control condition, the cell was stimulated repeatedly by identical zero-symmetrical white-noise motion. The mean firing rate was manipulated by adding a constant velocity offset either in the same area of the receptive field where the dynamic stimulus was displayed or in a separate one. We determined the information rate in the resulting spike trains in the time domain as the difference between the total and the noise entropy rate and found that the information rate increases with increasing mean firing under both stimulus conditions.     

On the existence of 'fast' and 'slow' directionally sensitive motion detector neurons in insects.

In a fly, butterfly, locust and dragonfly we examined the responses of a variety of directional motion-sensitive neurons which run from the brain down the ventral cord. The stimulus was a sinusoidally modulated moving pattern of regular stripes presented at a range of velocities in random order for either 0.1 s or 2.0 s. The response was measured as the total number of spikes to each stimulus. The neurons fall into two groups, 'fast' and 'slow'. The responses of the fast type rise progressively to a peak contrast frequency at 15-20 Hz for all four insects, and decline at higher contrast frequencies. The responses of slow neurons rise rapidly to a peak at 1-10 Hz and then decline more slowly across the range where the fast neurons are at their peak. The existence of two groups of neurons with overlapping response ranges to different velocities of the same pattern, presented in exactly the same way, provides the insect with a means of measuring angular velocity irrespective of contrast, spatial frequency or intensity. As an input mechanism it is proposed that there are two types of unit motion detector, fast and slow, the latter being the main input to the optomotor system. It is also argued that even these inputs are not sufficient to provide a mechanism for the whole repertoire of normal insect vision.     

Dscam and neuronal uniqueness

In the fruit fly Drosophila the gene encoding the cell adhesion molecule Dscam generates alternatively spliced mRNAs that can be translated into thousands of different protein isoforms. Three recent papers show that isoform-specific homophilic Dscam interactions cause dendritic branches of the same neuron to avoid each other (Hughes et al., 2007; Soba et al., 2007; Matthews et al., 2007). This process ensures the correct patterning of dendrites in the peripheral nervous system.     

Two classes of visual motion sensitive interneurons differ in direction and velocity dependency of in vivo calcium dynamics

Neurons exploit both membrane biophysics and biochemical pathways of the cytoplasm for dendritic integration of synaptic input. Here we quantify the tuning discrepancy of electrical and chemical response properties in two kinds of neurons using in vivo visual stimulation. Dendritic calcium concentration changes and membrane potential of visual interneurons of the fly were measured in response to visual motion stimuli. Two classes of tangential cells of the lobula plate were compared, HS-cells and CH-cells. Both neuronal classes are known to receive retinotopic input with similar properties, yet they differ in morphology, physiology, and computational context. Velocity tuning and directional selectivity of the electrical and calcium responses were investigated. In both cell classes, motion-induced calcium accumulation did not follow the early transient of the membrane potential. Rather, the amplitude of the calcium signal seemed to be related to the late component of the depolarization, where it was close to a steady state. Electrical and calcium responses differed with respect to their velocity tuning in CH-cells, but not in HS-cells. Furthermore, velocity tuning of the calcium response, but not of the electrical response differed between neuronal classes. While null-direction motion caused hyperpolarization in both classes, this led to a calcium decrement in CH-cells, but had no effect on the calcium signal in HS-cells, not even when calcium levels had been raised by a preceding excitatory motion stimulus. Finally, the voltage-[Ca2+]i-relationship for motion-induced, transient potential changes was steeper and less rectifying in CH-cells than in HS-cells. These results represent an example of dendritic information processing in vivo, where two neuronal classes respond to identical stimuli with a similar electrical response, but differing calcium response. This highlights the capacity of neurons to segregate two response components.     

A general model for visual motion detection

We propose a general model for detection of both first-order motion and second-order motion. In this model an input stimulus is divided into a number of partially overlapping spatiotemporal local regions. Spatiotemporal frequency analysis is done for every local region using Gabor filters, then the input stimulus (original spatiotemporal signal) is replaced by the outputs of Gabor filters. Local motion is detected by applying Gabor motion detectors to each local spatiotemporal pattern depicted by each local feature value. Outputs of all the detectors are integrated to give the final output for global motion of the input stimulus. The model was simulated on a computer and was confirmed to correctly detect second-order motion as well as first-order motion.     

Pattern-dependent response modulations in motion-sensitive visual interneurons--a model study

Even if a stimulus pattern moves at a constant velocity across the receptive field of motion-sensitive neurons, such as lobula plate tangential cells (LPTCs) of flies, the response amplitude modulates over time. The amplitude of these response modulations is related to local pattern properties of the moving retinal image. On the one hand, pattern-dependent response modulations have previously been interpreted as 'pattern-noise', because they deteriorate the neuron's ability to provide unambiguous velocity information. On the other hand, these modulations might also provide the system with valuable information about the textural properties of the environment. We analyzed the influence of the size and shape of receptive fields by simulations of four versions of LPTC models consisting of arrays of elementary motion detectors of the correlation type (EMDs). These models have previously been suggested to account for many aspects of LPTC response properties. Pattern-dependent response modulations decrease with an increasing number of EMDs included in the receptive field of the LPTC models, since spatial changes within the visual field are smoothed out by the summation of spatially displaced EMD responses. This effect depends on the shape of the receptive field, being the more pronounced--for a given total size--the more elongated the receptive field is along the direction of motion. Large elongated receptive fields improve the quality of velocity signals. However, if motion signals need to be localized the velocity coding is only poor but the signal provides--potentially useful--local pattern information. These modelling results suggest that motion vision by correlation type movement detectors is subject to uncertainty: you cannot obtain both an unambiguous and a localized velocity signal from the output of a single cell. Hence, the size and shape of receptive fields of motion sensitive neurons should be matched to their potential computational task.