Improved single pellet grasping using automated ad libitum full-time training robot

ReachingBot: an automated and scalable benchtop device for highly parallel Single Pellet Reach-and-Grasp training and assessment in mice

BACKGROUND

The single pellet reaching and grasp (SPRG) task is a behavioural assay widely used to study motor learning, control and recovery after nervous system injury in animals. The manual training and assessment of the SPRG is labour intensive and time consuming and has led to the development of multiple devices which automate the SPRG task.

NEW METHOD

Here, using robotics, computer vision, and machine learning analysis of videos, we describe a device that can be left unattended, presents pellets to mice, and, using two supervised learning algorithms, classifies the outcome of each trial with an accuracy of greater than 94% without the use of graphical processing units (GPUs). Our devices can also be operated using our cross-platform Graphical User Interface (GUI).

RESULTS

We show that these devices train and assess mice in parallel. 21 out of 30 mice retrieved >40% of pellets successfully following the training period. Following ischaemic stroke; some mice showed large persistent deficits whilst others showed only transient deficits. This highlights the heterogeneity in reaching outcomes following stroke.

COMPARISON WITH EXISTING METHOD(S)

Current state-of-the-art desktop methods either still require supervision, manual classification of trial outcome, or expensive locally-installed hardware such as graphical processing units (GPUs).

CONCLUSIONS

ReachingBots successfully automated SPRG training and assessment and revealed the heterogeneity in reaching outcomes following stroke. We conjecture that reach-and-grasp is represented in motor cortex bilaterally but with greater asymmetry in some mice than in others.

AutoRG: An automatized reach-to-grasp platform technology for assessing forelimb motor function, neural circuit activation, and cognition in rodents

BACKGROUND

Rodent reach-to-grasp function assessment is a translationally powerful model for evaluating neurological function impairments and recovery responses. Existing assessment platforms are experimenter-dependent, costly, or low-throughput with limited output measures. Further, a direct histologic comparison of neural activation has never been conducted between any novel, automated platform and the well-established single pellet skilled reach task (SRT).

NEW METHOD

To address these technological and knowledge gaps, we designed an open-source, low-cost Automatized Reach-to-Grasp (AutoRG) pull platform that reduces experimenter interventions and variability. We assessed reach-to-grasp function in rats across seven progressively difficult stages using AutoRG. We mapped AutoRG and SRT-activated motor circuitries in the rat brain using volumetric imaging of the immediate early gene-encoded Arc (activity-regulated cytoskeleton-associated) protein.

RESULTS

Rats demonstrated robust forelimb reaching and pulling behavior after training in AutoRG. Reliable force versus time responses were recorded for individual reach events in real time, which were used to derive several secondary functional measures of performance. Moreover, we provide the first demonstration that for a training period of 30 min, AutoRG and SRT both engage similar neural responses in the caudal forelimb area (CFA), rostral forelimb area (RFA), and sensorimotor area (S1).

CONCLUSION

AutoRG is the first low-cost, open-source pull system designed for the scale-up of volitional forelimb motor function testing and characterization of rodent reaching behavior. The similarities in neuronal activation patterns observed in the rat motor cortex after SRT and AutoRG assessments validate the AutoRG as a rigorously characterized, scalable alternative to the conventional SRT and expensive commercial systems.

A dystonia mouse model with motor and sequencing deficits paralleling human disease

The dystonias are a group of movement disorders characterized by involuntary twisting movements and postures. A lack of well characterized behavioral models of dystonia has impeded identification of circuit abnormalities giving rise to the disease. Most mouse behavioral assays are implemented independently of cortex, but cortical dysfunction is implicated in human dystonia. It is therefore important to identify dystonia models in which motor cortex-dependent behaviors are altered in ways relevant to human disease. The goal of this study was to characterize a cortically-dependent behavior in the recently-developed Dlx-CKO mouse model of DYT1 dystonia. Mice performed two tasks: skilled reaching and water-elicited grooming. These tests assess motor learning, dexterous skill, and innate motor sequencing. Furthermore, skilled reaching depends strongly on motor cortex, while dorsal striatum is critical for normal grooming. Dlx-CKO mice exhibited significantly lower success rates and pellet contacts compared to control mice during skilled reaching. Despite the skilled reaching impairments, Dlx-CKO mice adapt their reaching strategies. With training, they more consistently contacted the target. Grooming patterns of Dlx-CKO mice are more disorganized than in control mice, as evidenced by a higher proportion of non-chain grooming. However, when Dlx-CKO mice engage in syntactic chains, they execute them similarly to control mice. These abnormalities may provide targets for preclinical intervention trials, as well as facilitate determination of the physiologic path from torsinA dysfunction to motor phenotype.

When Spinal Neuromodulation Meets Sensorimotor Rehabilitation: Lessons Learned From Animal Models to Regain Manual Dexterity After a Spinal Cord Injury

Electrical neuromodulation has strongly hit the foundations of spinal cord injury and repair. Clinical and experimental studies have demonstrated the ability to neuromodulate and engage spinal cord circuits to recover volitional motor functions lost after the injury. Although the science and technology behind electrical neuromodulation has attracted much of the attention, it cannot be obviated that electrical stimulation must be applied concomitantly to sensorimotor rehabilitation, and one would be very difficult to understand without the other, as both need to be finely tuned to efficiently execute movements. The present review explores the difficulties faced by experimental and clinical neuroscientists when attempting to neuromodulate and rehabilitate manual dexterity in spinal cord injured subjects. From a translational point of view, we will describe the major rehabilitation interventions employed in animal research to promote recovery of forelimb motor function. On the other hand, we will outline some of the state-of-the-art findings when applying electrical neuromodulation to the spinal cord in animal models and human patients, highlighting how evidences from lumbar stimulation are paving the path to cervical neuromodulation.

Changes in reach-to-grasp behaviour over the course of training in rats

One complex task involving sequence of movements and movement refinement in the rat is the single-pellet reaching task, comprising orientation, transport and withdrawal in sequence. In turn, orientation comprises front wall detection, slot localization and nose poke until reach start. Video recordings of a rat in the reaching box highlighted three stages of temporal training: start of training (ST), forepaw dominance appearance (D) and fully trained (T). Regarding orientation, ST versus D and T presented a significant smaller frequency of approach to the front wall and a significant higher number of whisker cycles and nose touches during slot localization, involving a significant longer Orientation. At the ST stage, 44% of the trials were interrupted after nose poke, and poke took place at significant higher level from the shelf. The shelf was identified only when short whiskers contacted it, but the tongue and both forepaws were used without distinction to reach and grasp the pellet until a forepaw emerged as dominant at D stage. Regarding the temporal features of transport and withdrawal, comparing the D versus T stage revealed a significant longer duration. Finally, successes were significantly higher in T respect to D, meaning that after dominance emergence, more training was still necessary to improve reaching/grasping performance. This study provides evidence that, during training, the rats develop a strategy to obtain the pellets and then refine their movement pattern.

Automation of training and testing motor and related tasks in pre-clinical behavioural and rehabilitative neuroscience

Testing and training animals in motor and related tasks is a cornerstone of pre-clinical behavioural and rehabilitative neuroscience. Yet manually testing and training animals in these tasks is time consuming and analyses are often subjective. Consequently, there have been many recent advances in automating both the administration and analyses of animal behavioural training and testing. This review is an in-depth appraisal of the history of, and recent developments in, the automation of animal behavioural assays used in neuroscience. We describe the use of common locomotor and non-locomotor tasks used for motor training and testing before and after nervous system injury. This includes a discussion of how these tasks help us to understand the underlying mechanisms of neurological repair and the utility of some tasks for the delivery of rehabilitative training to enhance recovery. We propose two general approaches to automation: automating the physical administration of behavioural tasks (i.e., devices used to facilitate task training, rehabilitative training, and motor testing) and leveraging the use of machine learning in behaviour analysis to generate large volumes of unbiased and comprehensive data. The advantages and disadvantages of automating various motor tasks as well as the limitations of machine learning analyses are examined. In closing, we provide a critical appraisal of the current state of automation in animal behavioural neuroscience and a prospective on some of the advances in machine learning we believe will dramatically enhance the usefulness of these approaches for behavioural neuroscientists.

New Mechanistic Insights, Novel Treatment Paradigms, and Clinical Progress in Cerebrovascular Diseases

The past decade has brought tremendous progress in diagnostic and therapeutic options for cerebrovascular diseases as exemplified by the advent of thrombectomy in ischemic stroke, benefitting a steeply increasing number of stroke patients and potentially paving the way for a renaissance of neuroprotectants. Progress in basic science has been equally impressive. Based on a deeper understanding of pathomechanisms underlying cerebrovascular diseases, new therapeutic targets have been identified and novel treatment strategies such as pre- and post-conditioning methods were developed. Moreover, translationally relevant aspects are increasingly recognized in basic science studies, which is believed to increase their predictive value and the relevance of obtained findings for clinical application.This review reports key results from some of the most remarkable and encouraging achievements in neurovascular research that have been reported at the 10th International Symposium on Neuroprotection and Neurorepair. Basic science topics discussed herein focus on aspects such as neuroinflammation, extracellular vesicles, and the role of sex and age on stroke recovery. Translational reports highlighted endovascular techniques and targeted delivery methods, neurorehabilitation, advanced functional testing approaches for experimental studies, pre-and post-conditioning approaches as well as novel imaging and treatment strategies. Beyond ischemic stroke, particular emphasis was given on activities in the fields of traumatic brain injury and cerebral hemorrhage in which promising preclinical and clinical results have been reported. Although the number of neutral outcomes in clinical trials is still remarkably high when targeting cerebrovascular diseases, we begin to evidence stepwise but continuous progress towards novel treatment options. Advances in preclinical and translational research as reported herein are believed to have formed a solid foundation for this progress.

Self-directed rehabilitation training intensity thresholds for efficient recovery of skilled forelimb function in rats with cervical spinal cord injury

Task specific rehabilitation training is commonly used to treat motor dysfunction after neurological injures such as spinal cord injury (SCI), yet the use of task specific training in preclinical animal studies of SCI is not common. This is due in part to the difficulty in training animals to perform specific motor tasks, but also due to the lack of knowledge about optimal rehabilitation training parameters to maximize recovery. The single pellet reaching, grasping and retrieval (SPRGR) task (a.k.a. single pellet reaching task or Whishaw task) is a skilled forelimb motor task used to provide rehabilitation training and test motor recovery in rodents with cervical SCI. However, the relationships between the amount, duration, intensity, and timing of training remain poorly understood. In this study, using automated robots that allow rats with cervical SCI ad libitum access to self-directed SPRGR rehabilitation training, we show clear relationships between the total amount of rehabilitation training, the intensity of training (i.e., number of attempts/h), and performance in the task. Specifically, we found that rats naturally segregate into High and Low performance groups based on training strategy and performance in the task. Analysis of the different training strategies showed that more training (i.e., increased number of attempts in the SPRGR task throughout rehabilitation training) at higher intensities (i.e., number of attempts per hour) increased performance in the task, and that improved performance in the SPRGR task was linked to differences in corticospinal tract axon collateral densities in the injured spinal cords. Importantly, however, our data also indicate that rehabilitation training becomes progressively less efficient (i.e., less recovery for each attempt) as both the amount and intensity of rehabilitation training increases. Finally, we found that Low performing animals could increase their training intensity and transition to High performing animals in chronic SCI. These results highlight the rehabilitation training strategies that are most effective to regain skilled forelimb motor function after SCI, which will facilitate pre-clinical rehabilitation studies using animal models and could be beneficial in the development of more efficient clinical rehabilitation training strategies.

Behavioral testing in animal models of spinal cord injury

This review is based on a lecture presented at the Craig H. Neilsen Foundation sponsored Spinal Cord Injury Training Program at Ohio State University. We discuss the advantages and challenges of injury models in rodents and theory relation to various behavioral outcome measures. We offer strategies and advice on experimental design, behavioral testing, and on the challenges, one will encounter with animal testing. This review is designed to guide those entering the field of spinal cord injury and/or involved with in vivo animal testing.

The Home-Cage Automated Skilled Reaching Apparatus (HASRA): Individualized Training of Group-Housed Mice in a Single Pellet Reaching Task

The single pellet reaching task is commonly used in rodents to assess the acquisition of fine motor skill and recovery of function following nervous system injury. Although this task is useful for gauging skilled forelimb use in rodents, the process of training animals is labor intensive and variable across studies and labs. To address these limitations, we developed a single pellet reaching paradigm for training and testing group housed mice within their home cage. Mice enter a training compartment attached to the outside of the cage and retrieve millet seeds presented on a motorized pedestal that can be individually positioned to present seeds to either forelimb. To identify optimal training parameters, we compared task participation and success rates between groups of animals that were presented seeds at two different heights (floor vs mouth height) and at different intervals (fixed-time vs trial-based). The mouth height/fixed interval presentation style was most effective at promoting reaching behavior as all mice reached for seeds within 5 d. Using this paradigm, we assessed stroke-induced deficits in home-cage reaching. Following three weeks of baseline training, reaching success rate was ∼40%, with most trials performed during the dark cycle. A forelimb motor cortex stroke significantly decreased interaction with presented seeds within the first 2 d and impaired reaching success rates for the first 7 d. Our data demonstrate that group-housed mice can be efficiently trained on a single pellet reaching task in the home cage and that this assay is sensitive to stroke induced motor impairments.

Rehabilitative Training in Animal Models of Spinal Cord Injury

Rehabilitative motor training is currently one of the most widely used approaches to promote moderate recovery following injuries of the central nervous system. Such training is generally applied in the clinical setting, whereas it is not standard in preclinical research. This is a concern as it is becoming increasingly apparent that neuroplasticity enhancing treatments require training or some form of activity as a co-therapy to promote functional recovery. Despite the importance of training and the many open questions regarding its mechanistic consequences, its use in preclinical animal models is rather limited. Here we review approaches, findings and challenges when training is applied in animal models of spinal cord injury, and we suggest recommendations to facilitate the integration of training using an appropriate study design, into pre-clinical studies.

Automated Rat Single-Pellet Reaching with 3-Dimensional Reconstruction of Paw and Digit Trajectories

Rodent skilled reaching is commonly used to study dexterous skills, but requires significant time and effort to implement the task and analyze the behavior. Several automated versions of skilled reaching have been developed recently. Here, we describe a version that automatically presents pellets to rats while recording high-definition video from multiple angles at high frame rates (300 fps). The paw and individual digits are tracked with DeepLabCut, a machine learning algorithm for markerless pose estimation. This system can also be synchronized with physiological recordings, or be used to trigger physiologic interventions (e.g., electrical or optical stimulation).

Automated Forelimb Tasks for Rodents: Current Advantages and Limitations, and Future Promise

Rodent tests of function have advanced our understanding of movement, largely through the human training and testing and manual assessment. Tools such as reaching and grasping of a food pellet have been widely adopted because they are effective and simple to use. However, these tools are time-consuming, subjective, and often qualitative. Automation of training, testing, and assessment has the potential to increase efficiency while ensuring tasks are objective and quantitative. We detail new methods for automating rodent forelimb tests, including the use of pellet dispensers, sensors, computer vision, and home cage systems. We argue that limitations in existing forelimb tasks are driving the innovations in automated systems. We further argue that automated tasks partially address these limitations, and we outline necessary precautions and remaining challenges when adopting these types of tasks. Finally, we suggest attributes of future automated rodent assessment tools that can enable widespread adoption and help us better understand forelimb function in health and disease.

Sleep as a model to understand neuroplasticity and recovery after stroke: Observational, perturbational and interventional approaches

Our own experiences with disturbances to sleep demonstrate its crucial role in the recovery of cognitive functions. This importance is likely enhanced in the recovery from stroke; both in terms of its physiology and cognitive abilities. Decades of experimental research have highlighted which aspects and mechanisms of sleep are likely to underlie these forms of recovery. Conversely, damage to certain areas of the brain, as well as the indirect effects of stroke, may disrupt sleep. However, only limited research has been conducted which seeks to directly explore this bidirectional link between both the macro and micro-architecture of sleep and stroke. Here we describe a series of semi-independent approaches that aim to establish this link through observational, perturbational, and interventional experiments. Our primary aim is to describe the methodology for future clinical and translational research needed to delineate competing accounts of the current data. At the observational level we suggest the use of high-density EEG recording, combined analysis of macro and micro-architecture of sleep, detailed analysis of the stroke lesion, and sensitive measures of functional recovery. The perturbational approach attempts to find the causal links between sleep and stroke. We promote the use of transcranial magnetic stimulation combined with EEG to examine the cortical dynamics of the peri-infarct stroke area. Translational research should take this a step further using optogenetic techniques targeting more specific cell populations. The interventional approach focuses on how the same clinical and translational perturbational techniques can be adapted to influence long-term recovery of function.

The structure of arm and hand movements in a spontaneous and food rewarded on-line string-pulling task by the mouse

Arm and hand use by the mouse have been studied in a variety of tasks in order to understand the structure of skilled movements and motor learning, the anatomy and function of neural pathways, and to develop animal models of neurological conditions. The present study describes string-pulling by the mouse, a behavior in which a mouse uses hand-over-hand movements to pull down a string that hangs from the top of a test cage. Mice both spontaneously string-pull and also string-pull to obtain cashew nuts tied to the end of the string as food reward. To string-pull, mice sat upright and tracked the string with their nose and then made hand-over-hand movements to reel in the string. A string-pull movement consists of four arm movements (Advance to make purchase, Pull, Push to draw the string down and Lift to return the hand for the next Advance) and four hand movements (Collect to aim the hand, Overgrasp to position the hand, and Grasp to make purchase, and Release). The kinematic profiles of the string-pull movement are distinctive with each hand making similar movements at a rate of 4 cycles per second and with the Lift and Advance movements occurring at a higher speed than Pull and Push movements. The results are discussed in relation to the antecedent repertoire of mouse behavior that lends itself to string-pulling, with respect to the utility of using string-pulling to investigate motor systems and adapting string-pulling to model neurological conditions in mice.

Individualized tracking of self-directed motor learning in group-housed mice performing a skilled lever positioning task in the home cage

Skilled forelimb function in mice is traditionally studied through behavioral paradigms that require extensive training by investigators and are limited by the number of trials individual animals are able to perform within a supervised session. We developed a skilled lever positioning task that mice can perform within their home cage. The task requires mice to use their forelimb to precisely hold a lever mounted on a rotary encoder within a rewarded position to dispense a water reward. A Raspberry Pi microcomputer is used to record lever position during trials and to control task parameters, thus making this low-footprint apparatus ideal for use within animal housing facilities. Custom Python software automatically increments task difficulty by requiring a longer hold duration, or a more accurate hold position, to dispense a reward. The performance of individual animals within group-housed mice is tracked through radio-frequency identification implants, and data stored on the microcomputer may be accessed remotely through an active internet connection. Mice continuously engage in the task for over 2.5 mo and perform ~500 trials/24 h. Mice required ~15,000 trials to learn to hold the lever within a 10° range for 1.5 s and were able to further refine movement accuracy by limiting their error to a 5° range within each trial. These results demonstrate the feasibility of autonomously training group-housed mice on a forelimb motor task. This paradigm may be used in the future to assess functional recovery after injury or cortical reorganization induced by self-directed motor learning. NEW & NOTEWORTHY We developed a low-cost system for fully autonomous training of group-housed mice on a forelimb motor task. We demonstrate the feasibility of tracking both end-point, as well as kinematic performance of individual mice, with each performing thousands of trials over 2.5 mo. The task is run and controlled by a Raspberry Pi microcomputer, which allows for cages to be monitored remotely through an active internet connection.

A mouse's spontaneous eating repertoire aids performance on laboratory skilled reaching tasks: A motoric example of instinctual drift with an ethological description of the withdraw movements in freely-moving and head-fixed mice

Rodents display a spontaneous "order-common" pattern of food eating: they pick up food using the mouth, sit on their haunches, and transfer the food to the hands for handling/chewing. The present study examines how this pattern of behaviour influences performance on "skilled-reaching" tasks, in which mice purchase food with a single hand. Here five types of withdraw movement, the retraction of the hand, in three reaching tasks: freely-moving single-pellet, head-fixed single-pellet, and head-fixed pasta-eating is described. The withdraw movement varied depending upon whether a reach was anticipatory, no food present, or was unsuccessful or successful with food present. Ease of withdraw is dependent upon the extent to which animals used order-common movements. For freely-moving mice, a hand-to-mouth movement was assisted by a mouth-to-hand movement and food transfer to the mouth depended upon a sitting posture and using the other hand to assist food holding, both order-common movements. In the head-fixed single-pellet task, with postural and head movements prevented, withdraw was made with difficulty and tongue protrude movements assisted food transfer to the mouth once the hand reached the mouth. Only when a head-fixed mouse made a bilateral hand-to-mouth movement, a component of order-common eating, was the withdraw movement made with ease. The results are discussed with respect to the use of order-common movements in skilled-reaching tasks and with respect to the optimal design of tasks used to assess rodent skilled hand movement.

An Automated Home-Cage System to Assess Learning and Performance of a Skilled Motor Task in a Mouse Model of Huntington's Disease

Behavioral testing is a critical step in assessing the validity of rodent models of neurodegenerative disease, as well as evaluating the efficacy of pharmacological interventions. In models of Huntington's disease (HD), a gradual progression of impairments is observed across ages, increasing the need for sensitive, high-throughput and longitudinal assessments. Recently, a number of automated systems have been developed to perform behavioral profiling of animals within their own home-cage, allowing for 24-h monitoring and minimizing experimenter interaction. However, as of yet, few of these have had functionality for the assessment of skilled motor learning, a relevant behavior for movement disorders such as HD. To address this, we assess a lever positioning task within the mouse home-cage. Animals first acquire a simple operant response, before moving to a second phase where they must learn to hold the lever for progressively longer in a rewarded position range. Testing with this paradigm has revealed the presence of distinct phenotypes in the YAC128 mouse model of HD at three early symptomatic time points. YAC128 mice at two months old, but not older, had a motor learning deficit when required to adapt their response to changes in task requirements. In contrast, six-month-old YAC128 mice had disruptions of normal circadian activity and displayed kinematic abnormalities during performance of the task, suggesting an impairment in motor control. This system holds promise for facilitating high throughput behavioral assessment of HD mouse models for preclinical therapeutic screening.

An automated rat single pellet reaching system with high-speed video capture

BACKGROUND

Single pellet reaching is an established task for studying fine motor control in which rats reach for, grasp, and eat food pellets in a stereotyped sequence. Most incarnations of this task require constant attention, limiting the number of animals that can be tested and the number of trials per session. Automated versions allow more interventions in more animals, but must be robust and reproducible.

NEW METHOD

Our system automatically delivers single reward pellets for rats to grasp with their forepaw. Reaches are detected using real-time computer vision, which triggers video acquisition from multiple angles using mirrors. This allows us to record high-speed (>300 frames per second) video, and trigger interventions (e.g., optogenetics) with high temporal precision. Individual video frames are triggered by digital pulses that can be synchronized with behavior, experimental interventions, or recording devices (e.g., electrophysiology). The system is housed within a soundproof chamber with integrated lighting and ventilation, allowing multiple skilled reaching systems in one room.

RESULTS

We show that rats acquire the automated task similarly to manual versions, that the task is robust, and can be synchronized with optogenetic interventions.

COMPARISON WITH EXISTING METHODS

Existing skilled reaching protocols require high levels of investigator involvement, or, if ad libitum, do not allow for integration of high-speed, synchronized data collection.

CONCLUSION

This task will facilitate the study of motor learning and control by efficiently recording large numbers of skilled movements. It can be adapted for use with modern neurophysiology, which demands high temporal precision.

Single pellet grasping following cervical spinal cord injury in adult rat using an automated full-time training robot

Task specific motor training is a common form of rehabilitation therapy in individuals with spinal cord injury (SCI). The single pellet grasping (SPG) task is a skilled forelimb motor task used to evaluate recovery of forelimb function in rodent models of SCI. The task requires animals to obtain food pellets located on a shelf beyond a slit at the front of an enclosure. Manually training and testing rats in the SPG task requires extensive time and often yields results with high outcome variability and small therapeutic windows (i.e., the difference between pre- and post-SCI success rates). Recent advances in automated SPG training using automated pellet presentation (APP) systems allow rats to train ad libitum 24h a day, 7 days a week. APP trained rats have improved success rates, require less researcher time, and have lower outcome variability compared to manually trained rats. However, it is unclear whether APP trained rats can perform the SPG task using the APP system after SCI. Here we show that rats with cervical SCI can successfully perform the SPG task using the APP system. We found that SCI rats with APP training performed significantly more attempts, had slightly lower and less variable final score success rates, and larger therapeutic windows than SCI rats with manual training. These results demonstrate that APP training has clear advantages over manual training for evaluating reaching performance of SCI rats and represents a new tool for investigating rehabilitative motor training following CNS injury.