Biology subjects

Connelly, W. M.

Publications and source records attributed to Connelly, W. M..

8 recordsLinked to original sources

bioRxiv2026

A Smart Bait Dispenser with Real-time Monitoring for Species-Specific Bait Delivery

1. Remote image capture using trail cameras has become an essential part of many wildlife monitoring and management programs. Oral bait vaccination has been a valuable tool in wildlife disease management, but baits are generally dropped on the ground or delivered by devices that are typically developed to utilise physical and behavioural characteristics of the target species. Both approaches can result in a high proportion of baits being consumed by non-target species. Merging years of trail camera research knowledge with new technologies allows for development of more specific bait delivery options. 2. We developed an open-source smart bait dispenser that uses field-based image classification to selectively deliver baits to the target species. The system's long-range wide area network-based remote telemetry provides updates to an online dashboard in near real time to show camera triggers, target species detections, and baits dispensed. We trialled four dispensers with placebo baits for 62 nights across three field sites in Tasmania, Australia to deliver baits to Tasmanian devil (Sarcophilus harrisii). 3. The dispensers recorded 2106 motion triggers, of which 876 were triggered by animals. The onboard devil classification model achieved 98.6% accuracy, 75.0% precision, 77.4% recall, and a 76.2% F1-score. The dispenser released 41 placebo baits in total, of which 31 were target-specific distributions in response to presence of Tasmanian devils at the dispenser, hence achieving an overall bait delivery success rate of 75.6%. Importantly, no animal was able to forcibly retrieve baits from the dispenser. Observed average battery life across trials was 19.2 nights. 4. These results clearly demonstrate the effectiveness of our innovative bait dispenser for targeted bait delivery, offering a versatile ecological tool that can be tailored for various species by simply updating the image classification model. This adaptable system can support critical initiatives like wildlife oral bait vaccination and supplementary feeding after wildlife translocations.

ecology↗

bioRxiv2025

Feedforward and feedback population dynamics during binocular conflict in mouse visual cortex

Binocular rivalry arises when incongruent images are presented to the two eyes, producing stochastic alternations in perceptual dominance. While rivalry has been extensively studied in species with highly developed binocular vision, it is unclear whether similar representational dynamics occur in the mouse, a model system that allows large-scale cellular and circuit-level measurements. Here we used two-photon calcium imaging in awake mice to examine the population dynamics of primary visual cortex (V1) and long-range feedback axons from retrosplenial cortex (RSC) during presentation of dichoptically incongruent drifting gratings and natural movies. At the single-cell level, incongruent stimulation increased trial-to-trial variability of visually evoked responses relative to monocular stimulation. Linear SVM decoders trained on monocular responses revealed that during prolonged incongruent stimulation, V1 population activity alternated stochastically between representations of the two competing stimuli in a contrast-dependent manner. Decoder confidence was independent of pupil-indexed arousal state suggesting the dynamics observed may depend mostly on feedforward mechanisms. Transition analyses showed that switches in decoder output were typically driven by the emergence of responses to the ipsilateral stimulus, consistent with release from suppression of the non-dominant population. During incongruent presentation of natural movies, similar representational alternations were observed, indicating that rivalry-like dynamics were not dependent on orientation-selective adaptation. Imaging of RSC[->]V1 feedback axons revealed retinotopically specific, eye and orientation-selective signals that also alternated in dominance across time. These results establish the mouse as a model of rivalry-like cortical dynamics, demonstrate that both feedforward and feedback circuits contribute to representational alternation during binocular conflict, and provide a framework for mechanistic dissection of bistable perception.

neuroscience↗

bioRxiv2025

A study on edge devices for image classification of the Tasmanian devil (Sarcophilus harrisii) for vaccine delivery

O_LIA target-specific bait dispenser is required for oral bait vaccination of the endangered Tasmanian devil (Sarcophilus harrisii) against the deadly devil facial tumour disease. Development of a broadly extendible dispenser would be a beneficial ecological tool. A camera-based edge device with an onboard deep learning model can be used to identify the target species using image classification. It is important to choose a suitable edge device for the dispensers effective application in power constrained environments. C_LIO_LIWe evaluated four edge devices for the smart bait dispenser-ArduinoPro Nicla Vision, ArduinoPro Portenta H7 with Vision Shield (LoRa), Raspberry Pi 3 Model B+ and Raspberry Pi Zero 2 W. Two simple convolutional neural networks, and four fine-tuned pretrained models (MobileNetV2, MobileNetV3Small, ResNet50V2, and ResNet152V2) were trained on trail camera images of devil and non-devil species. These models were evaluated across four metrics and post-training quantised for deployment. The edge devices were assessed on inference latency of each model in seconds (s) and power consumption in watt (W). C_LIO_LIWe found that a simple CNN based image classification model yielded the best overall result of [~]96% across all metrics. Raspberry Pi 3 Model B+ and Zero 2 W could run all the six models whereas Portenta failed to run the two ResNet models and Nicla Vision failed to run all four pretrained models. All edge devices were found to be quick enough for dispenser application (average inference latency of 0.718 s). Portenta consumed the lowest power during inference (0.19 W), idle (0.185 W), and light sleep (0.045 W) states whereas Nicla Vision consumed the lowest power during deep sleep (0.002 W). C_LIO_LIOverall, we found that simple traditional CNNs was best suited for species classification in a camera-based smart bait dispenser for vaccination of Tasmanian devils. We conclude that ArduinoPro Portenta H7 with Vision Shield is the most competent device for this edge AI application. We suggest integration of this system with a microcontroller unit like ATtiny85 to minimise inactive power consumption of the dispenser. This method can be readily adapted for other species in vaccination and supplemental feeding projects. C_LI

ecology↗

bioRxiv2024

Cellular and circuit mechanisms underlying binocular vision

How binocular visual information is combined at the level of single neurons in the brain is not fully understood. Here, we show in mice that callosal input from the opposite visual cortex (V1) plays a critical role in this process. In vivo we find this callosal projection carries ipsilateral eye information and synapses exclusively onto binocular neurons. Using the presence of callosal input to identify binocular neurons in vitro, at the cellular level we show that binocular neurons are less excitable than monocular neurons due to high expression of Kv1 potassium channels. At the circuit level we find that only monocular neurons send callosal projections to the opposite V1, whereas binocular neurons do not. Finally, using dual-colour optogenetics we show that most binocular and monocular neurons receive direct input from the thalamus. In summary, we describe distinct cellular and circuit mechanisms underlying processing of binocular visual information in mouse V1.

neuroscience↗

bioRxiv2024

Gluk4-containing kainate receptors regulate synaptic communication in the motor cortex and reduce axon degeneration in adult mice

Glutamate-gated kainate receptors comprising the Gluk4 subunit (encoded by Grik4) are highly expressed by neurons in the central nervous system. We report that Grik4 mRNA is widely expressed by neurons in the adult mouse motor cortex, where GluK4-containing kainate receptors account for [~]60% of the kainate evoked current in layer V pyramidal neurons. To elucidate their role in motor circuit regulation, we analysed the behaviour of mice that lacked the pore forming domain of the GluK4 subunit (Grik4-/-mice). Grik4-/- mice were hyperactive, had an abnormal gait, and impaired motor coordination. At postnatal day (P)60, layer V pyramidal neurons received fewer miniature excitatory post synaptic currents, had a reduced density of thin spines on their basal dendrites, and a reduced density of VGlut1 puncta at the soma, consistent with neurons receiving fewer excitatory synaptic connections. Grik4-/- mice also lost [~]44% of their callosal axons between P60 and P180 and the amplitude of the callosal compound action potential was reduced by [~]25-30%. RNA sequencing data support the capacity for Grik4 to modulate synaptic and neuroprotective signalling pathways.

neuroscience↗

bioRxiv2021

Endogenous recruitment of frontal-sensory circuits during visual discrimination

A long-range circuit linking anterior cingulate cortex (ACC) to primary visual cortex (V1) has been previously proposed to mediate visual selective attention in mice during visually guided behaviour. Here we used in vivo two-photon functional imaging to measure endogenous activity of ACC neurons projecting to layer 1 of V1 (ACC-V1axons) in mice either passively viewing stimuli or performing a go/no-go visually guided task. We observed that while ACC-V1axons were recruited under these conditions, this was not linked to enhancement of neural or behavioural measures of sensory coding. Instead, ACC-V1axon activity was observed to be associated with licking behaviour, modulated by reward, and biased towards task relevant sensory cortex.

neuroscience↗

bioRxiv2020

RPG: A low-cost, open-source, high-performance solution for displaying visual stimuli

The development of new high throughput approaches for neuroscience such as high-density silicon probes and 2-photon imaging have led to a renaissance in visual neuroscience. However, generating the stimuli needed to evoke activity in the visual system still represents a non-negligible difficulty for experimentalists. While several widely used software toolkits exist to deliver such stimuli, they all suffer from some shortcomings. Primarily, the hardware needed to effectively display such stimuli comes at a significant financial cost, and secondly, triggering and/or timing the stimuli such that it can be accurately synchronized with other devices requires the use of legacy hardware, further hardware, or bespoke solutions. Here we present RPG, a Python package written for the Raspberry Pi, which overcomes these issues. Specifically, the Raspberry Pi is a low-cost, credit card sized computer with general purpose input/output pins, allowing RPG to be triggered to deliver stimuli and to provide real-time feedback on stimulus timing. RPG delivers stimuli at >60 frames per second and the feedback of frame timings is accurate to 10s of microseconds. We provide a simple to use Python interface that is capable of generating drifting sine wave gratings, Gabor patches and displaying raw images/video.

neuroscience↗

bioRxiv2019

Stable encoding of visual cues in the mouse retrosplenial cortex

The rodent retrosplenial cortex functions as an integrative hub for sensory and motor signals, serving roles in both navigation and memory. While retrosplenial cortex (RSC) is reciprocally connected with the sensory cortex, the form in which sensory information is represented in the retrosplenial cortex and how it interacts with behavioural state is unclear. Here, we used 2-photon cellular imaging of neural activity of putative excitatory (CaMKII expressing) and inhibitory (parvalbumin expressing) neurons to measure visual and running evoked activity in RSC and compare it to primary visual cortex (V1). We found that stimulus position and orientation information was preserved between V1 and RSC, and additionally that positional information was organised topographically. Stimulus directional preference was biased towards nasal-temporal flow. Locomotion modulation of activity of single neurons, both in darkness and light, was also more pronounced in RSC than V1, and strongest in parvalbumin-positive neurons. Longitudinal measurements of single neurons showed that these response features were stably maintained over many weeks. These data provide evidence for stable representations of visual cues in retrosplenial cortex which are highly spatially selective. These may provide sensory data to contribute to the formation of memories of spatial information.

neuroscience↗