Deep Learning Classification of Canine Behavior Using a Single Collar-Mounted Accelerometer: Real-World Validation

Mobility, functionality and functional mobility: A review and application for canine veterinary patients

Mobility is an essential aspect of a dog's daily life. It is defined as the ability to move freely and easily and deviations from an animals' normal mobility capabilities are often an indicator of disease, injury or pain. When a dog's mobility is compromised, often functionality (ability to perform activities of daily living [ADL]), is also impeded, which can diminish an animal's quality of life. Given this, it is necessary to understand the extent to which conditions impact a dog's physiological ability to move around their environment to carry out ADL, a concept termed functional mobility. In contrast to human medicine, validated measures of canine functional mobility are currently limited. The aim of this review is to summarise the extent to which canine mobility and functionality are associated with various diseases and how mobility and functional mobility are currently assessed within veterinary medicine. Future work should focus on developing a standardised method of assessing functional mobility in dogs, which can contextualise how a wide range of conditions impact a dog's daily life. However, for a true functional mobility assessment to be developed, a greater understanding of what activities dogs do on a daily basis and movements underpinning these activities must first be established.

Walk or be walked by the dog? The attachment role

BACKGROUND

The human-animal bond has been recognized as having positive effects on the health and well-being of both humans and pets. The present study aims to explore the influence of attachment on physical activity (PA), lifestyle, and health outcomes of dog owners (DO), highlighting the mutual benefits resulting from the relationship between DO and dogs.

METHODS

Thirty-eight DO and their dogs participated in this study. Socio-demographic data, the Self-Rated Health (SRH), FANTASTICO Lifestyle Scale, and the Lexington Attachment Pet Scale (LAPS) were assessed. PA was measured in both the DO and the dogs, using an ActiGraph GT3X accelerometer in the context of daily routine. Descriptive statistics and Spearman rank correlation analyses were performed to examine the associations between LAPS, PA levels, socio-demographic variables, lifestyle behaviors, and SRH.

RESULTS

Significant correlations were found between the dog owners' light-level PA and the pets' vigorous level of PA (rho = 0.445, p = 0.01). Furthermore, the importance of the pets' health (rho = -0.785, p = 0.02) and the LAPS subscales, namely proximity (rho = 0.358, p = 0.03), and attachment (rho = 0.392, p = 0.01), were related to taking the pet for a walk. Regarding lifestyle, DO with a healthier lifestyle had a better self-assessment of their health using the SRH (rho = 0.39, p = 0.02). Moreover, DO with better lifestyles also exhibited greater concern for their pet's health (rho = 0.398, p = 0.01).

CONCLUSIONS

This study emphasizes that individuals who adopt healthier habits tend to perceive themselves as healthier and exhibit greater concern for their pets' health. The attachment between DO and dogs is important in promoting healthy lifestyle behaviors and engagement in PA. Our results highlight that the presence of a dog is associated with a higher level of PA in DO, depending on the strength of the human-animal bond.

Feasibility of in-home electroencephalographic and actigraphy recordings in dogs

Introduction

Idiopathic epilepsy is a prevalent neurological disease in dogs. Dogs with epilepsy often present with behavioral comorbidities such as aggression, anxiety, and fear. These behaviors are consistent with pre, post, or interictal behaviors, prodromal changes, seizure-precipitating factors, or absence and focal seizures. The overlap in behavior presentations and lack of objective research methods for quantifying and classifying canine behavior makes determining the cause difficult. Behavioral comorbidities in addition to the task of caring for an epileptic animal have a significant negative impact on dog and caregiver quality of life.

Methods

This pilot study aimed to assess the feasibility of a novel technology combination for behavior classification and epileptic seizure detection for a minimum 24-h recording in the dog's home environment. It was expected that combining electroencephalography (EEG), actigraphy, and questionnaires would be feasible in the majority of trials. A convenience sample of 10 community-owned dogs was instrumented with wireless video-EEG and actigraphy for up to 48 h of recording at their caregiver's home. Three questionnaires (maximum 137 questions) were completed over the recording period by caregivers to describe their dog's everyday behavior and habits.

Results

Six of the 10 included dogs had combined EEG and actigraphy recordings for a minimum of 24 h.

Discussion

This shows that in-home EEG and actigraphy recordings are possible in community-owned dogs and provides a basis for a prospective study examining the same technology combination in a larger sample size.

Response of pet owners to Whistle FIT® activity monitor digital alerts of increased pruritic activity in their dogs: a retrospective observational study

Pruritus is a common clinical sign in dogs and is often underrecognized by dog owners and veterinarians. The Whistle FIT®, a wearable accelerometer paired with analytics, can detect changes in pruritic activity in dogs, which can be reported to owners in a smartphone/tablet application. The objectives of this retrospective observational study were to investigate the impact of digital alerts for increased pruritic behaviors received by dog owners in a real-life setting, on (1) the initiation of veterinary clinic visits, and (2) if such visits resulted in initiation of therapy for pruritus. Whistle FIT® data and electronic health records from 1,042 Banfield veterinary clinics in the United States were obtained for a 20-month period and reviewed retrospectively. Data on times of increased pruritic behaviors was calculated retrospectively by the investigators by applying the same algorithms used in the Whistle system. Data from the first 10-month interval was compared to the second 10 months, when reports on pruritic behaviors and alerts for increased pruritic behaviors were viewable by pet owners. Signalment of dogs with clinic visits in the first (n = 7,191) and second (n = 6,684) 10-month groups was similar. The total number of pruritic alerts was 113,530 in the first 10 months and 93,217 in the second 10 months. The odds of an 'alert visit' (the first veterinary clinic visit that occurred within 4 weeks after the time of a pruritus alert) was statistically significantly more likely (odds ratio, 1.6264; 95% CI, 1.57-1.69; p < 0.0001) in the second 10-month period compared to the first 10-month period. The total number of medications administered was 10,829 in the first 10 months and 9,863 in the second 10 months. The percentage of medications prescribed within 4 weeks after a pruritus alert was higher in the second 10 month period (53.3%) compared to the first 10 month period (38.8%). This study suggests that pruritus alerts sent to dog owners may improve owner recognition of pruritic behaviors and increase the likelihood of a veterinary visit to treat canine pruritus.

Machine learning prediction and classification of behavioral selection in a canine olfactory detection program

There is growing interest in canine behavioral research specifically for working dogs. Here we take advantage of a dataset of a Transportation Safety Administration olfactory detection cohort of 628 Labrador Retrievers to perform Machine Learning (ML) prediction and classification studies of behavioral traits and environmental effects. Data were available for four time points over a 12 month foster period after which dogs were accepted into a training program or eliminated. Three supervised ML algorithms had robust performance in correctly predicting which dogs would be accepted into the training program, but poor performance in distinguishing those that were eliminated (~ 25% of the cohort). The 12 month testing time point yielded the best ability to distinguish accepted and eliminated dogs (AUC = 0.68). Classification studies using Principal Components Analysis and Recursive Feature Elimination using Cross-Validation revealed the importance of olfaction and possession-related traits for an airport terminal search and retrieve test, and possession, confidence, and initiative traits for an environmental test. Our findings suggest which tests, environments, behavioral traits, and time course are most important for olfactory detection dog selection. We discuss how this approach can guide further research that encompasses cognitive and emotional, and social and environmental effects.

Machine learning based canine posture estimation using inertial data

The aim of this study was to design a new canine posture estimation system specifically for working dogs. The system was composed of Inertial Measurement Units (IMUs) that are commercially available, and a supervised learning algorithm which was developed for different behaviours. Three IMUs, each containing a 3-axis accelerometer, gyroscope, and magnetometer, were attached to the dogs' chest, back, and neck. To build and test the model, data were collected during a video-recorded behaviour test where the trainee assistance dogs performed static postures (standing, sitting, lying down) and dynamic activities (walking, body shake). Advanced feature extraction techniques were employed for the first time in this field, including statistical, temporal, and spectral methods. The most important features for posture prediction were chosen using Select K Best with ANOVA F-value. The individual contributions of each IMU, sensor, and feature type were analysed using Select K Best scores and Random Forest feature importance. Results showed that the back and chest IMUs were more important than the neck IMU, and the accelerometers were more important than the gyroscopes. The addition of IMUs to the chest and back of dog harnesses is recommended to improve performance. Additionally, statistical and temporal feature domains were more important than spectral feature domains. Three novel cascade arrangements of Random Forest and Isolation Forest were fitted to the dataset. The best classifier achieved an f1-macro of 0.83 and an f1-weighted of 0.90 for the prediction of the five postures, demonstrating a better performance than previous studies. These results were attributed to the data collection methodology (number of subjects and observations, multiple IMUs, use of common working dog breeds) and novel machine learning techniques (advanced feature extraction, feature selection and modelling arrangements) employed. The dataset and code used are publicly available on Mendeley Data and GitHub, respectively.

Margined Horn-Shaped Air Chamber for Body-Conduction Microphone

The sound amplification ratios of sealed air chambers with different shapes were quantitatively compared to design a body-conduction microphone to measure animal scratching sounds. Recently, quantitative monitoring of scratching intensity in dogs has been required. We have already developed a collar with a body-conduction microphone to measure body-conducted scratching sounds. However, the air chamber, one of the components of the body-conduction microphone, has not been appropriately designed. This study compared the amplification ratios of air chambers with different shapes through numerical analysis and experiments. According to the results, the horn-shaped air chamber achieved the highest amplification performance, at least for sound frequencies below 3 kHz. The simulated amplification ratio of the horn-shaped air chamber with a 1 mm height and a 15 mm diameter was 52.5 dB. The deformation of the bottom of the air chamber affected the amplification ratio. Adjusting the margin of the margined horn shape could maintain its amplification ratio at any pressing force. The simulated and experimental amplification ratios of the margined horn-shaped air chamber were 53.4 dB and 19.4 dB, respectively.

TN-GAN-Based Pet Behavior Prediction through Multiple-Dimension Time-Series Augmentation

Behavioral prediction modeling applies statistical techniques for classifying, recognizing, and predicting behavior using various data. However, performance deterioration and data bias problems occur in behavioral prediction. This study proposed that researchers conduct behavioral prediction using text-to-numeric generative adversarial network (TN-GAN)-based multidimensional time-series augmentation to minimize the data bias problem. The prediction model dataset in this study used nine-axis sensor data (accelerometer, gyroscope, and geomagnetic sensors). The ODROID N2+, a wearable pet device, collected and stored data on a web server. The interquartile range removed outliers, and data processing constructed a sequence as an input value for the predictive model. After using the z-score as a normalization method for sensor values, cubic spline interpolation was performed to identify the missing values. The experimental group assessed 10 dogs to identify nine behaviors. The behavioral prediction model used a hybrid convolutional neural network model to extract features and applied long short-term memory techniques to reflect time-series features. The actual and predicted values were evaluated using the performance evaluation index. The results of this study can assist in recognizing and predicting behavior and detecting abnormal behavior, capacities which can be applied to various pet monitoring systems.

Behavior-Based Video Summarization System for Dog Health and Welfare Monitoring

The popularity of dogs has been increasing owing to factors such as the physical and mental health benefits associated with raising them. While owners care about their dogs' health and welfare, it is difficult for them to assess these, and frequent veterinary checkups represent a growing financial burden. In this study, we propose a behavior-based video summarization and visualization system for monitoring a dog's behavioral patterns to help assess its health and welfare. The system proceeds in four modules: (1) a video data collection and preprocessing module; (2) an object detection-based module for retrieving image sequences where the dog is alone and cropping them to reduce background noise; (3) a dog behavior recognition module using two-stream EfficientNetV2 to extract appearance and motion features from the cropped images and their respective optical flow, followed by a long short-term memory (LSTM) model to recognize the dog's behaviors; and (4) a summarization and visualization module to provide effective visual summaries of the dog's location and behavior information to help assess and understand its health and welfare. The experimental results show that the system achieved an average F1 score of 0.955 for behavior recognition, with an execution time allowing real-time processing, while the summarization and visualization results demonstrate how the system can help owners assess and understand their dog's health and welfare.

Automatic gross tumor segmentation of canine head and neck cancer using deep learning and cross-species transfer learning

Background

Radiotherapy (RT) is increasingly being used on dogs with spontaneous head and neck cancer (HNC), which account for a large percentage of veterinary patients treated with RT. Accurate definition of the gross tumor volume (GTV) is a vital part of RT planning, ensuring adequate dose coverage of the tumor while limiting the radiation dose to surrounding tissues. Currently the GTV is contoured manually in medical images, which is a time-consuming and challenging task.

Purpose

The purpose of this study was to evaluate the applicability of deep learning-based automatic segmentation of the GTV in canine patients with HNC.

Materials and methods

Contrast-enhanced computed tomography (CT) images and corresponding manual GTV contours of 36 canine HNC patients and 197 human HNC patients were included. A 3D U-Net convolutional neural network (CNN) was trained to automatically segment the GTV in canine patients using two main approaches: (i) training models from scratch based solely on canine CT images, and (ii) using cross-species transfer learning where models were pretrained on CT images of human patients and then fine-tuned on CT images of canine patients. For the canine patients, automatic segmentations were assessed using the Dice similarity coefficient (Dice), the positive predictive value, the true positive rate, and surface distance metrics, calculated from a four-fold cross-validation strategy where each fold was used as a validation set and test set once in independent model runs.

Results

CNN models trained from scratch on canine data or by using transfer learning obtained mean test set Dice scores of 0.55 and 0.52, respectively, indicating acceptable auto-segmentations, similar to the mean Dice performances reported for CT-based automatic segmentation in human HNC studies. Automatic segmentation of nasal cavity tumors appeared particularly promising, resulting in mean test set Dice scores of 0.69 for both approaches.

Conclusion

In conclusion, deep learning-based automatic segmentation of the GTV using CNN models based on canine data only or a cross-species transfer learning approach shows promise for future application in RT of canine HNC patients.

Enhanced Classification of Dog Activities with Quaternion-Based Fusion Approach on High-Dimensional Raw Data from Wearable Sensors

The employment of machine learning algorithms to the data provided by wearable movement sensors is one of the most common methods to detect pets' behaviors and monitor their well-being. However, defining features that lead to highly accurate behavior classification is quite challenging. To address this problem, in this study we aim to classify six main dog activities (standing, walking, running, sitting, lying down, and resting) using high-dimensional sensor raw data. Data were received from the accelerometer and gyroscope sensors that are designed to be attached to the dog's smart costume. Once data are received, the module computes a quaternion value for each data point that provides handful features for classification. Next, to perform the classification, we used several supervised machine learning algorithms, such as the Gaussian naïve Bayes (GNB), Decision Tree (DT), K-nearest neighbor (KNN), and support vector machine (SVM). In order to evaluate the performance, we finally compared the proposed approach's F-score accuracies with the accuracy of classic approach performance, where sensors' data are collected without computing the quaternion value and directly utilized by the model. Overall, 18 dogs equipped with harnesses participated in the experiment. The results of the experiment show a significantly enhanced classification with the proposed approach. Among all the classifiers, the GNB classification model achieved the highest accuracy for dog behavior. The behaviors are classified with F-score accuracies of 0.94, 0.86, 0.94, 0.89, 0.95, and 1, respectively. Moreover, it has been observed that the GNB classifier achieved 93% accuracy on average with the dataset consisting of quaternion values. In contrast, it was only 88% when the model used the dataset from sensors' data.

FedAAR: A Novel Federated Learning Framework for Animal Activity Recognition with Wearable Sensors

Deep learning dominates automated animal activity recognition (AAR) tasks due to high performance on large-scale datasets. However, constructing centralised data across diverse farms raises data privacy issues. Federated learning (FL) provides a distributed learning solution to train a shared model by coordinating multiple farms (clients) without sharing their private data, whereas directly applying FL to AAR tasks often faces two challenges: client-drift during local training and local gradient conflicts during global aggregation. In this study, we develop a novel FL framework called FedAAR to achieve AAR with wearable sensors. Specifically, we devise a prototype-guided local update module to alleviate the client-drift issue, which introduces a global prototype as shared knowledge to force clients to learn consistent features. To reduce gradient conflicts between clients, we design a gradient-refinement-based aggregation module to eliminate conflicting components between local gradients during global aggregation, thereby improving agreement between clients. Experiments are conducted on a public dataset to verify FedAAR’s effectiveness, which consists of 87,621 two-second accelerometer and gyroscope data. The results demonstrate that FedAAR outperforms the state-of-the-art, on precision (75.23%), recall (75.17%), F1-score (74.70%), and accuracy (88.88%), respectively. The ablation experiments show FedAAR’s robustness against various factors (i.e., data sizes, communication frequency, and client numbers).

The unmet potential of artificial intelligence in veterinary medicine

Veterinary medicine is a broad and growing discipline that includes topics such as companion animal health, population medicine and zoonotic diseases, and agriculture. In this article, we provide insight on how artificial intelligence works and how it is currently applied in veterinary medicine. We also discuss its potential in veterinary medicine. Given the rapid pace of research and commercial product developments in this area, the next several years will pose challenges to understanding, interpreting, and adopting this powerful and evolving technology. Artificial intelligence has the potential to enable veterinarians to perform tasks more efficiently while providing new insights for the management and treatment of disorders. It is our hope that this will translate to better quality of life for animals and those who care for them.

Dog Behavior Recognition Based on Multimodal Data from a Camera and Wearable Device

Although various studies on monitoring dog behavior have been conducted, methods that can minimize or compensate data noise are required. This paper proposes multimodal data-based dog behavior recognition that fuses video and sensor data using a camera and a wearable device. The video data represent the moving area of dogs to detect the dogs. The sensor data represent the movement of the dogs and extract features that affect dog behavior recognition. Seven types of behavior recognition were conducted, and the results of the two data types were used to recognize the dog’s behavior through a fusion model based on deep learning. Experimentation determined that, among FasterRCNN, YOLOv3, and YOLOv4, the object detection rate and behavior recognition accuracy were the highest when YOLOv4 was used. In addition, the sensor data showed the best performance when all statistical features were selected. Finally, it was confirmed that the performance of multimodal data-based fusion models was improved over that of single data-based models and that the CNN-LSTM-based model had the best performance. The method presented in this study can be applied for dog treatment or health monitoring, and it is expected to provide a simple way to estimate the amount of activity.

Multi-level Hierarchical Complex Behavior Monitoring System for Dog Psychological Separation Anxiety Symptoms

An increasing number of people own dogs due to the emotional benefits they bring to their owners. However, many owners are forced to leave their dogs at home alone, increasing the risk of developing psychological disorders such as separation anxiety, typically accompanied by complex behavioral symptoms including excessive vocalization and destructive behavior. Hence, this work proposes a multi-level hierarchical early detection system for psychological Separation Anxiety (SA) symptoms detection that automatically monitors home-alone dogs starting from the most fundamental postures, followed by atomic behaviors, and then detecting separation anxiety-related complex behaviors. Stacked Long Short-Term Memory (LSTM) is utilized at the lowest level to recognize postures using time-series data from wearable sensors. Then, the recognized postures are input into a Complex Event Processing (CEP) engine that relies on knowledge rules employing fuzzy logic (Fuzzy-CEP) for atomic behaviors level and higher complex behaviors level identification. The proposed method is evaluated utilizing data collected from eight dogs recruited based on clinical inclusion criteria. The experimental results show that our system achieves approximately an F1-score of 0.86, proving its efficiency in separation anxiety symptomatic complex behavior monitoring of a home-alone dog.

Deep Learning Empowered Wearable-Based Behavior Recognition for Search and Rescue Dogs

Search and Rescue (SaR) dogs are important assets in the hands of first responders, as they have the ability to locate the victim even in cases where the vision and or the sound is limited, due to their inherent talents in olfactory and auditory senses. In this work, we propose a deep-learning-assisted implementation incorporating a wearable device, a base station, a mobile application, and a cloud-based infrastructure that can first monitor in real-time the activity, the audio signals, and the location of a SaR dog, and second, recognize and alert the rescuing team whenever the SaR dog spots a victim. For this purpose, we employed deep Convolutional Neural Networks (CNN) both for the activity recognition and the sound classification, which are trained using data from inertial sensors, such as 3-axial accelerometer and gyroscope and from the wearable’s microphone, respectively. The developed deep learning models were deployed on the wearable device, while the overall proposed implementation was validated in two discrete search and rescue scenarios, managing to successfully spot the victim (i.e., obtained F1-score more than 99%) and inform the rescue team in real-time for both scenarios.

Effect of Music on Stress Parameters in Dogs during a Mock Veterinary Visit

Simple Summary

Visits to the vet are stressful for many pet dogs, but less is known about how measures of stress change over the course of a visit. Identifying appropriate measures of canine stress, along with successful interventions which alleviate stress in dogs during a veterinary visit, will be of great benefit to dogs and people. Music therapy has been successfully used to reduce stress and anxiety in people and other animals. Specifically, a process called entrainment, which involves playing music at a particular tempo aimed at synchronizing physiological responses, has been implemented with success in humans. The aim of this study was to examine a range of behavioral and physiological measures in dogs over the duration of a veterinary visit and to establish if bespoke music, which mimicked the tempo of their resting heart rate, could improve wellbeing. The results indicated that certain measures increased over time, indicating that dogs became increasingly stressed. Music was not shown to have a demonstrated effect across measures, suggesting that the stressor may be too extreme for this type of intervention to have a positive effect, or that music therapy requires modification before it can be successful in alleviating stress in dogs during a veterinary visit.

Abstract

Veterinary visits can be stressful for dogs, but how their wellbeing changes during a visit is not well understood. Music therapy has been successfully used in clinical practice to alleviate stress and anxiety in people. The present study aimed to understand how canine stress changes during a veterinary visit, establish the effect of music, and highlight measures which may be of practical use. In a randomized crossover design, dogs were exposed to no music and a bespoke piece of classical music at a tempo designed to match their resting heart rate during a mock veterinary visit. Dogs were scored as more “afraid” during the physical examination compared to when they were in the hospital kennel (p < 0.001). Salivary cortisol, IgA, and infrared temperature all increased significantly (p < 0.05) from baseline to post-kennel and post-examination, with no effect of music treatment. Core body temperature (p = 0.010) and the odds of ‘relaxed’ lips (p = 0.020) were lower when dogs were exposed to music compared to control visits. Overall, dogs experienced changes in physiology and behavior, indicative of increased stress, over the course of the visit. Additional research is required to further understand the effect that bespoke music may have in alleviating canine stress during veterinary visits.

Cross-Modality Interaction Network for Equine Activity Recognition Using Imbalanced Multi-Modal Data

With the recent advances in deep learning, wearable sensors have increasingly been used in automated animal activity recognition. However, there are two major challenges in improving recognition performance—multi-modal feature fusion and imbalanced data modeling. In this study, to improve classification performance for equine activities while tackling these two challenges, we developed a cross-modality interaction network (CMI-Net) involving a dual convolution neural network architecture and a cross-modality interaction module (CMIM). The CMIM adaptively recalibrated the temporal- and axis-wise features in each modality by leveraging multi-modal information to achieve deep intermodality interaction. A class-balanced (CB) focal loss was adopted to supervise the training of CMI-Net to alleviate the class imbalance problem. Motion data was acquired from six neck-attached inertial measurement units from six horses. The CMI-Net was trained and verified with leave-one-out cross-validation. The results demonstrated that our CMI-Net outperformed the existing algorithms with high precision (79.74%), recall (79.57%), F1-score (79.02%), and accuracy (93.37%). The adoption of CB focal loss improved the performance of CMI-Net, with increases of 2.76%, 4.16%, and 3.92% in precision, recall, and F1-score, respectively. In conclusion, CMI-Net and CB focal loss effectively enhanced the equine activity classification performance using imbalanced multi-modal sensor data.