Automated Emotion Classification in Free-moving Rats

Studies involving emotion often use animal models and currently rely on manual labelling by researchers. This human-driven labelling approach leads to a number of challenges such as: long analysis times, imprecise results, observer drift, and varying correlation between observers. These problems impact reproducibility, and have contributed to our lack of understanding of fundamental mechanical questions such as how emotions arise from neuronal circuits. Recent success of machine learning models across similar problems show that it can help to mitigate these challenges while meeting or exceeding human accuracy. We developed a classifier pipeline that takes in videos and produces an emotion label. The pipeline extracts body part positions from each frame using a pose estimator and feeds them into an Artificial Neural Network (ANN) classifier built using stacked Long Short Term Memory (LSTM) layers. The data was collected by treating nine rats with Lypopolysaccharide (LPS) injections (10mg/kg). First, rats were recorded for 10 minutes under control conditions with no manipulation and no observed symptoms of stress or malaise. A week later, rats were injected with LPS and filmed for 10 minutes two hours post-injection. The classifier pipeline developed correctly labelled 78% of the 125,040 video segments from 8 test videos. When combined with a vote-based system, this led to 7 of the 8 test videos being classified correctly which was the same accuracy attained by a human expert from the lab. The test videos had varying environments and used rats that were different from the training videos, providing evidence of a degree of robustness in the model. Future work will focus on expanding the test data and incorporating models for 3D pose estimation and behavioral classification.

Estimation of aboveground herbaceous biomass using visually ranked digital photographs

Methods for estimating aboveground herbaceous biomass in the field have generally involved calibrating visual estimates against clipped, dried and weighed biomass samples, requiring lengthy periods of estimation and destructive sampling in the field. Here we developed and tested a photographic estimation technique (PET) that minimises field time and provides accurate estimates of aboveground herbaceous biomass. Photographs of the biomass to be estimated taken in the field are ranked against calibration images of known biomass in the laboratory. The study was conducted in New South Wales, Australia, in grassy forest dells and grasslands at Booroolong Nature Reserve in the temperate New England Tablelands Bioregion and in semi-arid grassy shrubland on Naree Station in the arid Mulga Lands Bioregion. Photographs of quadrats containing the herbaceous biomass to be estimated were taken in successive years at both sites. Calibration and validation quadrats were also photographed, and the vegetation clipped, bagged, dried and weighed. The calibration and validation photographs were rank-ordered independently by three observers in terms of estimated dry weight, and the validation quadrats assigned a putative dry weight by reference to the known weights of the calibration images in the rank order. The accuracy of each observer’s estimates was assessed by regressing the estimated weight of each validation quadrat against the actual weight, which was withheld from the observer during the estimation procedure. Regression analysis of visually estimated weights on actual weights of validation quadrats yielded regression coefficients (R2) of 0.80–0.98 and 0.81–0.97 in the temperate-zone and arid-zone sites, respectively. PET was reliably used to visually estimate aboveground herbaceous biomass across a range of communities in two different climatic zones. The benefits of PET include reduced field time, minimisation of destructive sampling and avoidance of observer drift in estimating biomass in the field.

Embracing Our Visual Inspection and Analysis Tradition

Most investigators using single-case experimental designs use interobserver agreement (IOA) checks to enhance the credibility of the collected data, and they report the results of those assessments using percentage of agreement estimates. An alternative is to graph both observers’ records of the measured behavior on the primary study graphs. Such graphing leads to greater transparency and is advocated for five reasons: (a) to make explicit how IOA assessments were distributed across the study, (b) to ensure agreement estimates are reported at the level of the measured behavior of interest rather than a broader observational code, (c) to detect observer drift, (d) to detect the effect of observer expectations, and (e) to put the IOA data in a more suitable context for assessing the internal validity of the study by eliminating the need for an arbitrary agreement criterion.

Direct Observation: Factors Affecting the Accuracy of Observers

The use of direct observation methods to collect data relevant to research and practice in special education is widespread. Although the reliability of such data has often been addressed, far less attention has focused on the accuracy of these data. The purposes of this article are (a) to review research on factors that adversely affect the accuracy of observers, and (b) to provide recommendations for reducing their possible influence. The areas discussed include reactivity, observer drift; the recording procedure; location of the observation; reliability; expectancy and feedback; and the characteristics of subjects, observers, and settings.