BigNeuron: a resource to benchmark and predict performance of algorithms for automated tracing of neurons in light microscopy datasets.

BigNeuron is an open community bench-testing platform with the goal of setting open standards for accurate and fast automatic neuron tracing. We gathered a diverse set of image volumes across several species that is representative of the data obtained in many neuroscience laboratories interested in neuron tracing. Here, we report generated gold standard manual annotations for a subset of the available imaging datasets and quantified tracing quality for 35 automatic tracing algorithms. The goal of generating such a hand-curated diverse dataset is to advance the development of tracing algorithms and enable generalizable benchmarking. Together with image quality features, we pooled the data in an interactive web application that enables users and developers to perform principal component analysis, t-distributed stochastic neighbor embedding, correlation and clustering, visualization of imaging and tracing data, and benchmarking of automatic tracing algorithms in user-defined data subsets. The image quality metrics explain most of the variance in the data, followed by neuromorphological features related to neuron size. We observed that diverse algorithms can provide complementary information to obtain accurate results and developed a method to iteratively combine methods and generate consensus reconstructions. The consensus trees obtained provide estimates of the neuron structure ground truth that typically outperform single algorithms in noisy datasets. However, specific algorithms may outperform the consensus tree strategy in specific imaging conditions. Finally, to aid users in predicting the most accurate automatic tracing results without manual annotations for comparison, we used support vector machine regression to predict reconstruction quality given an image volume and a set of automatic tracings.

Hoverfly (Eristalis tenax) pursuit of artificial targets.

The ability to visualize small moving objects is vital for the survival of many animals, as these could represent predators or prey. For example, predatory insects, including dragonflies, robber flies and killer flies, perform elegant, high-speed pursuits of both biological and artificial targets. Many non-predatory insects, including male hoverflies and blowflies, also pursue targets during territorial or courtship interactions. To date, most hoverfly pursuits have been studied outdoors. To investigate hoverfly (Eristalis tenax) pursuits under more controlled settings, we constructed an indoor arena that was large enough to encourage naturalistic behavior. We presented artificial beads of different sizes, moving at different speeds, and filmed pursuits with two cameras, allowing subsequent 3D reconstruction of the hoverfly and bead position as a function of time. We show that male E. tenax hoverflies are unlikely to use strict heuristic rules based on angular size or speed to determine when to start pursuit, at least in our indoor setting. We found that hoverflies pursued faster beads when the trajectory involved flying downwards towards the bead. Furthermore, we show that target pursuit behavior can be broken down into two stages. In the first stage, the hoverfly attempts to rapidly decreases the distance to the target by intercepting it at high speed. During the second stage, the hoverfly's forward speed is correlated with the speed of the bead, so that the hoverfly remains close, but without catching it. This may be similar to dragonfly shadowing behavior, previously coined 'motion camouflage'.

Avoiding obstacles while intercepting a moving target: a miniature fly's solution.

The miniature robber fly Holcocephala fusca intercepts its targets with behaviour that is approximated by the proportional navigation guidance law. During predatory trials, we challenged the interception of H. fusca performance by placing a large object in its potential flight path. In response, H. fusca deviated from the path predicted by pure proportional navigation, but in many cases still eventually contacted the target. We show that such flight deviations can be explained as the output of two competing navigational systems: pure-proportional navigation and a simple obstacle avoidance algorithm. Obstacle avoidance by H. fusca is here described by a simple feedback loop that uses the visual expansion of the approaching obstacle to mediate the magnitude of the turning-away response. We name the integration of this steering law with proportional navigation 'combined guidance'. The results demonstrate that predatory intent does not operate a monopoly on the fly's steering when attacking a target, and that simple guidance combinations can explain obstacle avoidance during interceptive tasks.

Medium compensation in a spring-actuated system.

Mantis shrimp strikes are one of the fastest animal movements, despite their occurrence in a water medium with viscous drag. Since the strike is produced by a latch-mediated spring-actuated system and not directly driven by muscle action, we predicted that strikes performed in air would be faster than underwater as a result of reduction in the medium's drag. Using high-speed video analysis of stereotyped strikes elicited from Squilla mantis, we found the exact opposite: strikes are much slower and less powerful in air than in water. S. mantis strikes in air have a similar mass and performance to latch-mediated spring-actuated jumps in locusts, suggesting a potential threshold for the energetics of a 1-2 g limb rotating in air. Drag forces induced by the media may be a key feature in the evolution of mantis shrimp strikes and provide a potential target for probing the braking system of these extremely fast movements.

Integration of Small- and Wide-Field Visual Features in Target-Selective Descending Neurons of both Predatory and Nonpredatory Dipterans.

For many animals, target motion carries high ecological significance as this may be generated by a predator, prey, or potential mate. Indeed, animals whose survival depends on early target detection are often equipped with a sharply tuned visual system, yielding robust performance in challenging conditions. For example, many fast-flying insects use visual cues for identifying targets, such as prey (e.g., predatory dragonflies and robberflies) or conspecifics (e.g., nonpredatory hoverflies), and can often do so against self-generated background optic flow. Supporting these behaviors, the optic lobes of insects that pursue targets harbor neurons that respond robustly to the motion of small moving objects, even when displayed against syn-directional background clutter. However, in diptera, the encoding of target information by the descending neurons, which are more directly involved in generating the behavioral output, has received less attention. We characterized target-selective neurons by recording in the ventral nerve cord of male and female predatory Holcocephala fusca robberflies and of male nonpredatory Eristalis tenax hoverflies. We show that both species have dipteran target-selective descending neurons that only respond to target motion if the background is stationary or moving slowly, moves in the opposite direction, or has un-naturalistic spatial characteristics. The response to the target is suppressed when background and target move at similar velocities, which is strikingly different to the response of target neurons in the optic lobes. As the neurons we recorded from are premotor, our findings affect our interpretation of the neurophysiology underlying target-tracking behaviors.SIGNIFICANCE STATEMENT Many animals use sensory cues to detect moving targets that may represent predators, prey, or conspecifics. For example, birds of prey show superb sensitivity to the motion of small prey, and intercept these at high speeds. In a similar manner, predatory insects visually track moving prey, often against cluttered backgrounds. Accompanying this behavior, the brains of insects that pursue targets contain neurons that respond exclusively to target motion. We here show that dipteran insects also have target-selective descending neurons in the part of their nervous system that corresponds to the vertebrate spinal cord. Surprisingly, and in contrast to the neurons in the brain, these premotor neurons are inhibited by background patterns moving in the same direction as the target.

Visual approach computation in feeding hoverflies.

On warm sunny days, female hoverflies are often observed feeding from a wide range of wild and cultivated flowers. In doing so, hoverflies serve a vital role as alternative pollinators, and are suggested to be the most important pollinators after bees and bumblebees. Unless the flower hoverflies are feeding from is large, they do not readily share the space with other insects, but instead opt to leave if another insect approaches. We used high-speed videography followed by 3D reconstruction of flight trajectories to quantify how female Eristalis hoverflies respond to approaching bees, wasps and two different hoverfly species. We found that, in 94% of the interactions, the occupant female left the flower when approached by another insect. We found that compared with spontaneous take-offs, the occupant hoverfly's escape response was performed at ∼3 times higher speed (spontaneous take-off at 0.2±0.05 m s-1 compared with 0.55±0.08 m s-1 when approached by another Eristalis). The hoverflies tended to take off upward and forward, while taking the incomer's approach angle into account. Intriguingly, we found that, when approached by wasps, the occupant Eristalis took off at a higher speed and when the wasp was further away. This suggests that feeding hoverflies may be able to distinguish these predators, demanding impressive visual capabilities. Our results, including quantification of the visual information available before occupant take-off, provide important insight into how freely behaving hoverflies perform escape responses from competitors and predators (e.g. wasps) in the wild.

Eight pairs of descending visual neurons in the dragonfly give wing motor centers accurate population vector of prey direction.

Intercepting a moving object requires prediction of its future location. This complex task has been solved by dragonflies, who intercept their prey in midair with a 95% success rate. In this study, we show that a group of 16 neurons, called target-selective descending neurons (TSDNs), code a population vector that reflects the direction of the target with high accuracy and reliability across 360°. The TSDN spatial (receptive field) and temporal (latency) properties matched the area of the retina where the prey is focused and the reaction time, respectively, during predatory flights. The directional tuning curves and morphological traits (3D tracings) for each TSDN type were consistent among animals, but spike rates were not. Our results emphasize that a successful neural circuit for target tracking and interception can be achieved with few neurons and that in dragonflies this information is relayed from the brain to the wing motor centers in population vector form.

The Killer Fly Hunger Games: Target Size and Speed Predict Decision to Pursuit.

Predatory animals have evolved to optimally detect their prey using exquisite sensory systems such as vision, olfaction and hearing. It may not be so surprising that vertebrates, with large central nervous systems, excel at predatory behaviors. More striking is the fact that many tiny insects, with their miniscule brains and scaled down nerve cords, are also ferocious, highly successful predators. For predation, it is important to determine whether a prey is suitable before initiating pursuit. This is paramount since pursuing a prey that is too large to capture, subdue or dispatch will generate a substantial metabolic cost (in the form of muscle output) without any chance of metabolic gain (in the form of food). In addition, during all pursuits, the predator breaks its potential camouflage and thus runs the risk of becoming prey itself. Many insects use their eyes to initially detect and subsequently pursue prey. Dragonflies, which are extremely efficient predators, therefore have huge eyes with relatively high spatial resolution that allow efficient prey size estimation before initiating pursuit. However, much smaller insects, such as killer flies, also visualize and successfully pursue prey. This is an impressive behavior since the small size of the killer fly naturally limits the neural capacity and also the spatial resolution provided by the compound eye. Despite this, we here show that killer flies efficiently pursue natural (Drosophila melanogaster) and artificial (beads) prey. The natural pursuits are initiated at a distance of 7.9 ± 2.9 cm, which we show is too far away to allow for distance estimation using binocular disparities. Moreover, we show that rather than estimating absolute prey size prior to launching the attack, as dragonflies do, killer flies attack with high probability when the ratio of the prey's subtended retinal velocity and retinal size is 0.37. We also show that killer flies will respond to a stimulus of an angular size that is smaller than that of the photoreceptor acceptance angle, and that the predatory response is strongly modulated by the metabolic state. Our data thus provide an exciting example of a loosely designed matched filter to Drosophila, but one which will still generate successful pursuits of other suitable prey.

Compound eyes and retinal information processing in miniature dipteran species match their specific ecological demands.

The compound eye of insects imposes a tradeoff between resolution and sensitivity, which should exacerbate with diminishing eye size. Tiny lenses are thought to deliver poor acuity because of diffraction; nevertheless, miniature insects have visual systems that allow a myriad of lifestyles. Here, we investigate whether size constraints result in an archetypal eye design shared between miniature dipterans by comparing the visual performance of the fruit fly Drosophila and the killer fly Coenosia. These closely related species have neural superposition eyes and similar body lengths (3 to 4 mm), but Coenosia is a diurnal aerial predator, whereas slow-flying Drosophila is most active at dawn and dusk. Using in vivo intracellular recordings and EM, we report unique adaptations in the form and function of their photoreceptors that are reflective of their distinct lifestyles. We find that although these species have similar lenses and optical properties, Coenosia photoreceptors have three- to fourfold higher spatial resolution and rate of information transfer than Drosophila. The higher performance in Coenosia mostly results from dramatically diminished light sensors, or rhabdomeres, which reduce pixel size and optical cross-talk between photoreceptors and incorporate accelerated phototransduction reactions. Furthermore, we identify local specializations in the Coenosia eye, consistent with an acute zone and its predatory lifestyle. These results demonstrate how the flexible architecture of miniature compound eyes can evolve to match information processing with ecological demands.

A distance-field based automatic neuron tracing method.

BACKGROUND: Automatic 3D digital reconstruction (tracing) of neurons embedded in noisy microscopic images is challenging, especially when the cell morphology is complex. RESULTS: We have developed a novel approach, named DF-Tracing, to tackle this challenge. This method first extracts the neurite signal (foreground) from a noisy image by using anisotropic filtering and automated thresholding. Then, DF-Tracing executes a coupled distance-field (DF) algorithm on the extracted foreground neurite signal and reconstructs the neuron morphology automatically. Two distance-transform based "force" fields are used: one for "pressure", which is the distance transform field of foreground pixels (voxels) to the background, and another for "thrust", which is the distance transform field of the foreground pixels to an automatically determined seed point. The coupling of these two force fields can "push" a "rolling ball" quickly along the skeleton of a neuron, reconstructing the 3D cell morphology. CONCLUSION: We have used DF-Tracing to reconstruct the intricate neuron structures found in noisy image stacks, obtained with 3D laser microscopy, of dragonfly thoracic ganglia. Compared to several previous methods, DF-Tracing produces better reconstructions.

Predictive saccades and decision making in the beetle-predating saffron robber fly.

Internal predictions about the sensory consequences of self-motion, encoded by corollary discharge, are ubiquitous in the animal kingdom, including for fruit flies, dragonflies, and humans. In contrast, predicting the future location of an independently moving external target requires an internal model. With the use of internal models for predictive gaze control, vertebrate predatory species compensate for their sluggish visual systems and long sensorimotor latencies. This ability is crucial for the timely and accurate decisions that underpin a successful attack. Here, we directly demonstrate that the robber fly Laphria saffrana, a specialized beetle predator, also uses predictive gaze control when head tracking potential prey. Laphria uses this predictive ability to perform the difficult categorization and perceptual decision task of differentiating a beetle from other flying insects with a low spatial resolution retina. Specifically, we show that (1) this predictive behavior is part of a saccade-and-fixate strategy, (2) the relative target angular position and velocity, acquired during fixation, inform the subsequent predictive saccade, and (3) the predictive saccade provides Laphria with additional fixation time to sample the frequency of the prey's specular wing reflections. We also demonstrate that Laphria uses such wing reflections as a proxy for the wingbeat frequency of the potential prey and that consecutively flashing LEDs to produce apparent motion elicits attacks when the LED flicker frequency matches that of the beetle's wingbeat cycle.

Evolution of visual system specialization in predatory arthropods.

Under strong selective pressure for survival, image-forming vision set off an ongoing predatory arms race 500 million years ago. Since then, and particularly so in the arthropods, predatory behavior has driven a myriad of eye adaptations that increase visual performance. In this review, we provide examples of how different arthropod predators have achieved improvements in key visual features such as spatial and temporal resolution of their retina. We then describe morphological, neural and behavioral strategies used by animals in this group to gather crucial information about the prey, such as its distance, velocity and size. We also highlight the importance of head and body tracking movements to aid in categorizing the potential prey, and briefly mention the ongoing work on the sensorimotor transformations necessary for target interception.

Overexpressing temperature-sensitive dynamin decelerates phototransduction and bundles microtubules in Drosophila photoreceptors.

shibire(ts1), a temperature-sensitive mutation of the Drosophila gene encoding a Dynamin orthologue, blocks vesicle endocytosis and thus synaptic transmission, at elevated, or restrictive temperatures. By targeted Gal4 expression, UAS-shibire(ts1) has been used to dissect neuronal circuits. We investigated the effects of UAS-shibire(ts1) overexpression in Drosophila photoreceptors at permissive (19 degrees C) and restrictive (31 degrees C) temperatures. At 19 degrees C, overexpression of UAS-shi(ts1) causes decelerated phototransduction and reduced neurotransmitter release. This phenotype is exacerbated with dark adaptation, age and in white mutants. Photoreceptors overexpressing UAS-shibire(ts1) contain terminals with widespread vacuolated mitochondria, reduced numbers of vesicles and bundled microtubules. Immuno-electron microscopy reveals that the latter are dynamin coated. Further, the microtubule phenotype is not restricted to photoreceptors, as UAS-shibire(ts1) overexpression in lamina cells also bundles microtubules. We conclude that dynamin has multiple functions that are interrupted by UAS-shibire(ts1) overexpression in Drosophila photoreceptors, destabilizing their neural communication irreversibly at previously reported permissive temperatures.

Binocular Encoding in the Damselfly Pre-motor Target Tracking System.

Akin to all damselflies, Calopteryx (family Calopterygidae), commonly known as jewel wings or demoiselles, possess dichoptic (separated) eyes with overlapping visual fields of view. In contrast, many dragonfly species possess holoptic (dorsally fused) eyes with limited binocular overlap. We have here compared the neuronal correlates of target tracking between damselfly and dragonfly sister lineages and linked these changes in visual overlap to pre-motor neural adaptations. Although dragonflies attack prey dorsally, we show that demoiselles attack prey frontally. We identify demoiselle target-selective descending neurons (TSDNs) with matching frontal visual receptive fields, anatomically and functionally homologous to the dorsally positioned dragonfly TSDNs. By manipulating visual input using eyepatches and prisms, we show that moving target information at the pre-motor level depends on binocular summation in demoiselles. Consequently, demoiselles encode directional information in a binocularly fused frame of reference such that information of a target moving toward the midline in the left eye is fused with information of the target moving away from the midline in the right eye. This contrasts with dragonfly TSDNs, where receptive fields possess a sharp midline boundary, confining responses to a single visual hemifield in a sagittal frame of reference (i.e., relative to the midline). Our results indicate that, although TSDNs are conserved across Odonata, their neural inputs, and thus the upstream organization of the target tracking system, differ significantly and match divergence in eye design and predatory strategies. VIDEO ABSTRACT.

Neural Control of Dynamic 3-Dimensional Skin Papillae for Cuttlefish Camouflage.

The color and pattern changing abilities of octopus, squid, and cuttlefish via chromatophore neuromuscular organs are unparalleled. Cuttlefish and octopuses also have a unique muscular hydrostat system in their skin. When this system is expressed, dermal bumps called papillae disrupt body shape and imitate the fine texture of surrounding objects, yet the control system is unknown. Here we report for papillae: (1) the motoneurons and the neurotransmitters that control activation and relaxation, (2) a physiologically fast expression and retraction system, and (3) a complex of smooth and striated muscles that enables long-term expression of papillae through sustained tension in the absence of neural input. The neural circuits controlling acute shape-shifting skin papillae in cuttlefish show homology to the iridescence circuits in squids. The sustained tension in papillary muscles for long-term camouflage utilizes muscle heterogeneity and points toward the existence of a "catch-like" mechanism that would reduce the necessary energy expenditure.

Interception by two predatory fly species is explained by a proportional navigation feedback controller.

When aiming to capture a fast-moving target, animals can follow it until they catch up, or try to intercept it. In principle, interception is the more complicated strategy, but also more energy efficient. To study whether simple feedback controllers can explain interception behaviours by animals with miniature brains, we have reconstructed and studied the predatory flights of the robber fly Holcocephala fusca and killer fly Coenosia attenuata Although both species catch other aerial arthropods out of the air, Holcocephala contrasts prey against the open sky, while Coenosia hunts against clutter and at much closer range. Thus, their solutions to this target catching task may differ significantly. We reconstructed in three dimensions the flight trajectories of these two species and those of the presented targets they were attempting to intercept. We then tested their recorded performances against simulations. We found that both species intercept targets on near time-optimal courses. To investigate the guidance laws that could underlie this behaviour, we tested three alternative control systems (pure pursuit, deviated pursuit and proportional navigation). Only proportional navigation explains the timing and magnitude of fly steering responses, but with differing gain constants and delays for each fly species. Holcocephala uses a dimensionless navigational constant of N ≈ 3 with a time delay of ≈28 ms to intercept targets over a comparatively long range. This constant is optimal, as it minimizes the control effort required to hit the target. In contrast, Coenosia uses a constant of N ≈ 1.5 with a time delay of ≈18 ms, this setting may allow Coenosia to cope with the extremely high line-of-sight rotation rates, which are due to close target proximity, and thus prevent overcompensation of steering. This is the first clear evidence of interception supported by proportional navigation in insects. This work also demonstrates how by setting different gains and delays, the same simple feedback controller can yield the necessary performance in two different environments.

A Novel Interception Strategy in a Miniature Robber Fly with Extreme Visual Acuity.

Our visual system allows us to rapidly identify and intercept a moving object. When this object is far away, we base the trajectory on the target's location relative to an external frame of reference [1]. This process forms the basis for the constant bearing angle (CBA) model, a reactive strategy that ensures interception since the bearing angle, formed between the line joining pursuer and target (called the range vector) and an external reference line, is held constant [2-4]. The CBA model may be a fundamental and widespread strategy, as it is also known to explain the interception trajectories of bats and fish [5, 6]. Here, we show that the aerial attack of the tiny robber fly Holcocephala fusca is consistent with the CBA model. In addition, Holcocephala fusca displays a novel proactive strategy, termed "lock-on" phase, embedded with the later part of the flight. We found the object detection threshold for this species to be 0.13°, enabled by an extremely specialized, forward pointing fovea (∼5 ommatidia wide, interommatidial angle Δφ = 0.28°, photoreceptor acceptance angle Δρ = 0.27°). This study furthers our understanding of the accurate performance that a miniature brain can achieve in highly demanding sensorimotor tasks and suggests the presence of equivalent mechanisms for target interception across a wide range of taxa. VIDEO ABSTRACT.

Target detection in insects: optical, neural and behavioral optimizations.

Motion vision provides important cues for many tasks. Flying insects, for example, may pursue small, fast moving targets for mating or feeding purposes, even when these are detected against self-generated optic flow. Since insects are small, with size-constrained eyes and brains, they have evolved to optimize their optical, neural and behavioral target visualization solutions. Indeed, even if evolutionarily distant insects display different pursuit strategies, target neuron physiology is strikingly similar. Furthermore, the coarse spatial resolution of the insect compound eye might actually be beneficial when it comes to detection of moving targets. In conclusion, tiny insects show higher than expected performance in target visualization tasks.

Virtual finger boosts three-dimensional imaging and microsurgery as well as terabyte volume image visualization and analysis.

Three-dimensional (3D) bioimaging, visualization and data analysis are in strong need of powerful 3D exploration techniques. We develop virtual finger (VF) to generate 3D curves, points and regions-of-interest in the 3D space of a volumetric image with a single finger operation, such as a computer mouse stroke, or click or zoom from the 2D-projection plane of an image as visualized with a computer. VF provides efficient methods for acquisition, visualization and analysis of 3D images for roundworm, fruitfly, dragonfly, mouse, rat and human. Specifically, VF enables instant 3D optical zoom-in imaging, 3D free-form optical microsurgery, and 3D visualization and annotation of terabytes of whole-brain image volumes. VF also leads to orders of magnitude better efficiency of automated 3D reconstruction of neurons and similar biostructures over our previous systems. We use VF to generate from images of 1,107 Drosophila GAL4 lines a projectome of a Drosophila brain.