Rowland Sillito

Semi-supervised Learning for Anomalous Trajectory Detection

A novel learning framework is proposed for anomalous behaviour detection in a video surveillance scenario, so that a classifier which distinguishes between normal and anomalous behaviour patterns can be incrementally trained with the assistance of a human operator. We consider the behaviour of pedestrians in terms of motion trajectories, and parametrise these trajectories using the control points of approximating cubic spline curves. This paper demonstrates an incremental semi-supervised one-class learning procedure in which unlabelled trajectories are combined with occasional examples of normal behaviour labelled by a human operator. This procedure is found to be effective on two different datasets, indicating that a human operator could potentially train the system to detect anomalous behaviour by providing only occasional interventions (a small percentage of the total number of observations).

Analysis of individual mouse activity in group housed animals of different inbred strains using a novel automated home cage analysis system.

Central nervous system disorders such as autism as well as the range of neurodegenerative diseases such as Huntingtons disease are commonly investigated using genetically altered mouse models. The current system for characterizing these mice usually involves removing the animals from their home-cage environment and placing them into novel environments where they undergo a battery of tests measuring a range of behavioral and physical phenotypes. These tests are often only conducted for short periods of times in social isolation. However, human manifestations of such disorders are often characterized by multiple phenotypes, presented over long periods of time and leading to significant social impacts. Here, we have developed a system which will allow the automated monitoring of individual mice housed socially in the cage they are reared and housed in, within established social groups and over long periods of time. We demonstrate that the system accurately reports individual locomotor behavior within the group and that the measurements taken can provide unique insights into the effects of genetic background on individual and group behavior not previously recognized.

Incremental one-class learning with bounded computational complexity

An incremental one-class learning algorithm is proposed for the purpose of outlier detection. Outliers are identified by estimating - and thresholding - the probability distribution of the training data. In the early stages of training a non-parametric estimate of the training data distribution is obtained using kernel density estimation. Once the number of training examples reaches the maximum computationally feasible limit for kernel density estimation, we treat the kernel density estimate as a maximally-complex Gaussian mixture model, and keep the model complexity constant bymerging a pair of components for each newkernel added. This method is shown to outperform a current state-of-the-art incremental one-class learning algorithm (Incremental SVDD [5]) on a variety of datasets, while requiring only an upper limit on model complexity to be specified.

Parametric Trajectory Representations for Behaviour Classification

This paper presents an empirical comparison of strategies for representing motion trajectories with fixed-length vectors. We compare four techniques, which have all previously been adopted in the trajectory classification literature: least-squares cubic spline approximation, the Discrete Fourier Transform, Chebyshev polynomial approximation, and the Haar wavelet transform. We measure the class separability of five different trajectory datasets - ranging from vehicle trajectories to pen trajectories - when described in terms of these representations. Results obtained over a range of dimensionalities indicate that the different representations yield similar levels of class separability, with marginal improvements provided by Chebyshev and Spline representations. For the datasets considered here, each representation appears to yield better results when used in conjunction with a curve parametrisation strategy based on arc-length, rather than time. However, we illustrate a situation - pertinent to surveillance applications - where the converse is true.

Assessing mouse behaviour throughout the light/dark cycle using automated in-cage analysis tools

Highlights • Automated assessment of mouse home-cage behaviour is robust and reliable.• Analysis over multiple light/dark cycles improves ability to classify behaviours.• Combined RFID and video analysis enables home-cage analysis in group housed animals.

Automated recording of home cage activity and temperature of individual rats housed in social groups: The Rodent Big Brother project

Measuring the activity and temperature of rats is commonly required in biomedical research. Conventional approaches necessitate single housing, which affects their behavior and wellbeing. We have used a subcutaneous radiofrequency identification (RFID) transponder to measure ambulatory activity and temperature of individual rats when group-housed in conventional, rack-mounted home cages. The transponder location and temperature is detected by a matrix of antennae in a baseplate under the cage. An infrared high-definition camera acquires side-view video of the cage and also enables automated detection of vertical activity. Validation studies showed that baseplate-derived ambulatory activity correlated well with manual tracking and with side-view whole-cage video pixel movement. This technology enables individual behavioral and temperature data to be acquired continuously from group-housed rats in their familiar, home cage environment. We demonstrate its ability to reliably detect naturally occurring behavioral effects, extending beyond the capabilities of routine observational tests and conventional monitoring equipment. It has numerous potential applications including safety pharmacology, toxicology, circadian biology, disease models and drug discovery.

MEK Inhibitors Reverse cAMP-Mediated Anxiety in Zebrafish

Summary Altered phosphodiesterase (PDE)-cyclic AMP (cAMP) activity is frequently associated with anxiety disorders, but current therapies act by reducing neuronal excitability rather than targeting PDE-cAMP-mediated signaling pathways. Here, we report the novel repositioning of anti-cancer MEK inhibitors as anxiolytics in a zebrafish model of anxiety-like behaviors. PDE inhibitors or activators of adenylate cyclase cause behaviors consistent with anxiety in larvae and adult zebrafish. Small-molecule screening identifies MEK inhibitors as potent suppressors of cAMP anxiety behaviors in both larvae and adult zebrafish, while causing no anxiolytic behavioral effects on their own. The mechanism underlying cAMP-induced anxiety is via crosstalk to activation of the RAS-MAPK signaling pathway. We propose that targeting crosstalk signaling pathways can be an effective strategy for mental health disorders, and advance the repositioning of MEK inhibitors as behavior stabilizers in the context of increased cAMP.

Pharmacological validation of individual animal locomotion, temperature and behavioural analysis in group-housed rats using a novel automated home cage analysis system: A comparison with the modified Irwin test

BACKGROUND The ActualHCA™ system continuously monitors the activity, temperature and behavior of group-housed rats without invasive surgery. The system was validated to detect the contrasting effects of sedative and stimulant test agents (chlorpromazine, clonidine and amphetamine), and compared with the modified Irwin test (mIT) with rectal temperature measurements. METHODS Six male Han Wistar rats per group were used to assess each test agent and vehicle controls in separate ActualHCA™ recordings and mIT. The mIT was undertaken at 15, 30 mins, 1, 2, 4 and 24 h post-dose. ActualHCA™ recorded continuously for 24 h post-dose under 3 experimental conditions: dosed during light phase, dark phase, and light phase with a scheduled cage change at the time of peak effects determined by mIT. RESULTS ActualHCA™ detected an increase stimulated activity from the cage change at 1-2 h post-dose which was obliterated by chlorpromazine and clonidine. Amphetamine increased activity up to 4 h post-dose in all conditions. Temperature from ActualHCA™ was affected by all test agents in all conditions. The mIT showed effects on all 3 test agents up to 4 h post-dose, with maximal effects at 1-2 h post-dose. The maximal effects on temperature from ActualHCA™ differed from mIT. Delayed effects on activity were detected by ActualHCA™, but not on mIT. CONCLUSIONS Continuous monitoring has the advantage of capturing effects over time that may be missed with manual tests using pre-determined time points. This automated behavioural system does not replace the need for conventional methods but could be implemented simultaneously to improve our understanding of behavioural pharmacology.

Rodent Big Brother: A Comparison to the Modified Irwin Test for Assessing Drug-Induced Changes in Activity and Temperature in Rates

Drug-Induced Changed in Activity and Temperature in Rats Karen Tse1, Amy Keerie1, Rowland Sillito2, Rachel Collier1, Claire Grant3, Catherine Vickers4, Kathryn Chapman4, J Douglas Armstrong2, Will S Redfern1 AstraZeneca, Fleming Building, Babraham Institute, Cambridge, CB22 3AT1, Actual Analytics Ltd, Wilkie Building, Edinburgh, EH8 9AG2, AstraZeneca, Alderley Park, Cheshire, SK10 4TG3, NC3Rs, Gibbs Building, 215 Euston Road, London, NW1 2BE4