Dalarna University's logo and link to the university's website

Advancements in modern medical technology have enabled cures for a fraction of patients while extending survival times for those who are not cured. For non-cured patients, disease recurrence is influenced by observed covariates and unobserved individual heterogeneity (random effects). In biomedical studies, dependent censoring is frequently encountered, for example, in cancer patients, where right censoring can be caused by death from unrelated diseases or due to an (unobservable) cure status. This study introduces a joint frailty model for recurrent event data with a cure fraction, effectively capturing heterogeneity and inducing dependent censoring. The proposed multivariate joint frailty mixture cure models incorporate covariates and frailties, together with the event incidence time and latent cure status. The model accounts for the probability of a cure after each recurrence using both the complementary log-log and the logistic link function. A likelihood-based estimation method is developed using the Monte Carlo Expectation-Maximization (MCEM) algorithm. Through Monte Carlo simulation, we examine the finite sample properties of the MCEM estimators, supplemented by a real-world application using secondary data on hospital readmissions for colorectal cancer recurrence post-surgery. Simulation results suggest lifetime and frailty parameter estimates are unbiased and consistent. Compared to models with identical frailty structure, both the complementary log-log and the logistic cure frailty models with dependent frailties demonstrate a better fit with the real data, as evidenced by lower Akaike information criteria values.

The structure of cubicles can hinder cows’ movements when transitioning between postures, leading to atypical motion patterns. Assessing posture transitions relies on visual observations. This study presents a framework for complementing these assessments with kinematic measurements using 3D pose estimation. A total 809 rising and 791 lying down posture transitions were recorded over 12 cubicles by 7 synchronized cameras and processed with 3D pose estimation locating the position of the poll, withers, T13 and sacrum. First, the displacement of the keypoints was used to detect phases of the posture transitions. This detection was compared with visual observations of 200 recordings. The average mean absolute difference in detected timestamps between human and machine across all phases was 0.5 s (average σ = 0.7) and was under 0.9 s for all phases. Second, indicators were scored based on spatial use and duration, and their distribution compared to existing thresholds. We observed that 59.9 % of rising bouts and 29.1 % of lying down bouts exceeded at least one threshold. Rising delay occurred in 2.8 % of rising bouts and backwards crawling in 59.2 %. Lying down duration exceeded the threshold in 28.9 % of bouts, and rear limbs shifting duration in 8.3 %. Side lunge had a binary threshold which was not adapted to continuous sensor data. Finally, we investigated the association between indicators and found distinct dimensions for head lunge and crawling. We conclude that 3D pose is useful to score posture transition indicators, and that several indicators should be used together to capture distinct dimensions. © 2025 Elsevier B.V., All rights reserved.

Access to mechanical brushes enables grooming behaviour in dairy cows and has shown benefits for cow welfare, including improved cleanliness, comfort, stress reduction. Brush-use may also promote a positive emotional state. Reduced brush use has been associated with health issues, suggesting its potential for automated health monitoring. This study aimed at evaluating whether data generated by pose estimation could be used to assess brush use patterns in loose-housed dairy cows. It presents an approach for automatically identifying the body segment being brushed as an application of pose estimation. Data collection was carried out at the Swedish Livestock Research Centre in a loose housing system equipped with an automatic milking system and two mechanical rotating brushes. Recordings spanned 25:30 h and used three cameras, at different positions, monitoring a single mechanical brush placed in a passageway between cubicle rows. One human observer with access to recordings from all three synchronized cameras annotated the data-set on a second-by-second basis. The observer recorded: (1) the number of cows using the brush; (2) the anatomical segment being brushed; and (3) whether brushing resumed after a pause. The same video recordings were processed with object detection and pose estimation, which predicted the location of bounding boxes for cows and for the brush as well as corresponding keypoints. Using the brush and cow keypoint locations, we attempted to detect brushing by anatomical region. In a first stage, machine-learning models were trained to predict brushing state (independent of location) using keypoint distance to the brush, achieving an accuracy of 86.3 %. To mitigate the risk of error propagation, we relied on human annotations to segment the video to confirmed brushing bouts for analysis in the second stage. To identify the anatomical location of brushing, two methods were evaluated: (1) simply assigning the brushing location to the closest keypoint, achieving 73 % average accuracy across classes, and (2) projecting brush and anatomical keypoints onto a spline modelling the cow's backline, resulting in 87 % accuracy. Misclassifications were predominantly limited to adjacent body segments. Given that intra-observer reliability was 90 %, the spline-based method was deemed sufficiently reliable for research applications to accurately monitor the specific body segments being brushed. © 2025 Elsevier B.V., All rights reserved.

AIM: Older adults are increasingly encouraged to continue living in their own homes with support from home care services. However, few studies have focused on older adults' safety in home care. This study explored associations between the sense of security and factors related to demographic characteristics and home care services.

METHODS: The mixed longitudinal design was based on a retrospective national survey. The study population consisted of individuals in Sweden (aged 65+ years) granted home care services at any time between 2016 and 2020 (n=82,834-94,714). Multiple ordinal logistic regression models were fitted using the generalised estimation equation method to assess the strength of relationship between the dependent (sense of security) and independent (demographics, health and care-related factors) variables.

RESULTS: The sense of security tended to increase between 2016 and 2020, and was significantly associated with being a woman, living outside big cities, being granted more home care services hours or being diagnosed/treated for depression (cumulative odds ratio 2-9% higher). Anxiety, poor health and living alone were most strongly associated with insecurity (cumulative odds ratio 17-64% lower). Aside from overall satisfaction with home care services, accessibility and confidence in staff influenced the sense of security most.

CONCLUSIONS: We stress the need to promote older adults' sense of security for safe ageing in place, as mandated by Swedish law. Home care services profoundly influence older adults' sense of security. Therefore, it is vital to prioritise continuity in care, establish trust and build relationships with older adults. Given the increasing shortage of staff, integrating complementary measures, such as welfare technologies, is crucial to promoting this sense of security.

Freestall comfort is reflected in various indicators, including the ability for dairy cattle to display unhindered posture transition movements in the cubicles. To ensure farm animal welfare, it is instrumental for the farm management to be able to continuously monitor occurrences of abnormal motions. Advances in computer vision have enabled accurate kinematic measurements in several fields such as human, equine and bovine biomechanics. An important step upstream to measuring displacement during posture transitions is to determine that the behavior is accurately detected. In this study, we propose a framework for detecting lying to standing posture transitions from 3D pose estimation data. A multi-view computer vision system recorded posture transitions between Dec. 2021 and Apr. 2022 in a Swedish stall housing 183 individual cows. The output data consisted of the 3D coordinates of specific anatomical landmarks. Sensitivity of posture transition detection was 88.2% while precision reached 99.5%. Analyzing those transition movements, breakpoints detected the timestamp of onset of the rising motion, which was compared with that annotated by observers. Agreement between observers, measured by intra-class correlation, was 0.85 between 3 human observers and 0.81 when adding the automated detection. The intra-observer mean absolute difference in annotated timestamps ranged from 0.4s to 0.7s. The mean absolute difference between each observer and the automated detection ranged from 1.0s to 1.3s. There was a significant difference in annotated timestamp between all observer pairs but not between the observers and the automated detection, leading to the conclusion that the automated detection does not introduce a distinct bias. We conclude that the model is able to accurately detect the phenomenon of interest and that it is equatable to an observer.

This study investigates the combination of audio and image data to classify road conditions, particularly focusingon loose gravel scenarios. The dataset underwent binary categorisation, comprising audio segments capturinggravel sounds and corresponding images. Early feature fusion, utilising a pre-trained Very Deep ConvolutionalNetworks 19 (VGG19) and Principal component analysis (PCA), improved the accuracy of the Random Forestclassifier, surpassing other models in accuracy, precision, recall, and F1-score. Late fusion, involving decisionlevelprocessing with logical disjunction and conjunction gates (AND and OR) in combination with individualclassifiers for images and audio based on Densely Connected Convolutional Networks 121 (DenseNet121),demonstrated notable performance, especially with the OR gate, achieving 97 % accuracy. The late fusionmethod enhances adaptability by compensating for limitations in one modality with information from the other.Adapting maintenance based on identified road conditions minimises unnecessary environmental impact. Thismethod can help to identify loose gravel on gravel roads, substantially improving road safety and implementing aprecise maintenance strategy through a data-driven approach.

Maintaining gravel roads is crucial, as loose gravel poses safety risks and increases vehicle costs. Current methods used by the Swedish road administration, Trafikverket, are subjective and time-consuming. Road agencies need a cost-effective, efficient, and unbiased approach to assess gravel road conditions. Studies show human ratings are error-prone and inconsistent. This study aims to develop an automatic method for estimating loose gravel using audio recordings from inside a vehicle, capturing the sound of gravel hitting the car's bottom. These recordings were classified into four classes based on Trafikverket regulations. Sound features were extracted and analysed using supervised machine-learning methods. The Multilayer Perceptron (MLP) achieved the highest classification accuracy of 0.96, with an F1 score, recall, and precision of 0.97. Results indicate that audio data can effectively classify loose gravel conditions.

The rigid structure of free stall partitions interferes with the natural movements of cows as they transition between postures, often resulting in abnormal motions like sideways head lunge. Although most stalls allow for forward lunge room, cows are frequently hindered in their posture transition movements. Methods are needed to evaluate effective lunge room and detect sideways lunge. The Sony multi-camera system generated 3D pose estimation to measure lunge distance and angle during lying-to-standing posture transitions (n = 493 bouts). After validating lunge timestamp detection against 3 observers (n = 165 annotated bouts), we explored features associated with abnormal rising. Agreement between observers on lunge timestamp was very high, as per an intra-class correlation of 0.97, and 0.95 when adding the automated detection. Mean absolute difference (MAD) in annotated lunge timestamp between observers and machine was 0.33s while mean difference was 0.03s ± 0.03 indicating a minor difference and no substantial bias. In comparison, average intra-observer MAD was 0.2s. Lunge angle had a mean of 166.1° ± 0.5 and was skewed to the left by -1.31, indicating that most motions occurred with the body in a relatively straight line and that sideways lunging occurred at a lower frequency. Using a linear regression, a significant effect of height at the withers (p = 0.003) and of lunge distance (P < 0.001) were found on lunge angle. These results show that the standardized cubicle does not accommodate all individuals' proper lunge. They further suggest that 3D pose estimation is a promising technology for measuring the kinematics of lunging motions. © 2024 11th European Conference on Precision Livestock Farming. All rights reserved.

The color of road signs is a critical aspect of road safety, as it helps drivers quickly and accurately identify and respond to these signs. Properly colored road signs improve visibility during the day and make it easier for drivers to make informed decisions while driving. In order to ensure the safety and efficiency of road traffic, it is essential to maintain the appropriate color level of road signs. The objective of this study was to analyze the color status and daylight chromaticity of in-use road signs using supervised machine learning models, and to explore the correlation between road sign's age and daylight chromaticity. Three algorithms were employed: Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN). The data used in this study was collected from road signs that were in-use on roads in Sweden. The study employed classification models to assess the color status (accepted or rejected) of the road signs based on minimum acceptable color levels according to standards, and regression models to predict the daylight chromaticity values. The correlation between road sign's age and daylight chromaticity was explored through regression analysis. Daylight chromaticity describes the color quality of road signs in daylight, that is expressed in terms of X and Y chromaticity coordinates. The study revealed a linear relationship between the road sign's age and daylight chromaticity for blue, green, red, and white sheeting, but not for yellow. The lifespan of red signs was estimated to be around 12 years, much shorter than the estimated lifespans of yellow, green, blue, and white sheeting, which are 35, 42, 45, and 75 years, respectively. The supervised machine learning models successfully assessed the color status of the road signs and predicted the daylight chromaticity values using the three algorithms. The results of this study showed that the ANN classification and ANN regression models achieved high accuracy of 81% and R2 of 97%, respectively. The RF and SVM models also performed well, with accuracy values of 74% and 79% and R2 ranging from 59% to 92%. The findings demonstrate the potential of machine learning to effectively predict the status and daylight chromaticity of road signs and their impact on road safety in the Swedish context. © 2023 International Association of Traffic and Safety Sciences