Livestock production benefits from automated monitoring of animal health and welfare. However, continuous tracking of commercial pigs using Computer Vision (CV) is hindered by their uniform appearance, the need to avoid reliance on physical tags in commercial settings, and changes in appearance due to growth and dirt. This study addresses longitudinal tracking through an open-set Re-Identification (ReID) framework based on deep metric learning, designed to maintain the identities of visually uniform group-housed pigs without artificial or natural marks. A ResNet50 backbone was fine-tuned with a combined ArcFace and semi-hard triplet loss on a large-scale longitudinal dataset of visually uniform pigs. To mitigate embedding drift, a multi-anchor voting algorithm matched daily cluster centroids against a rolling bank of historical identity representations, while custom synthetic dirt augmentation prevented overfitting to transient dirt patterns. Three modalities were evaluated: Instance Segmentation (IS), Oriented Bounding Boxes (OBB) and Axis-Aligned Bounding Boxes (AABB). During an 18-day open-set evaluation, the multi-anchor framework reached a mean tracking accuracy of 88.0% with IS images, outperforming a sequential day-to-day baseline (54.6%). OBB performance was comparable to IS at lower computational cost, whereas AABB degraded to 51.4%, attributed to background noise. Without synthetic dirt augmentation, tracking fell to 56.5% for IS images. An adapted GradCAM verified that the network differentiated individuals using stable anatomical traits, such as the head, shoulder and rear, rather than transient artefacts. These findings indicate that ReID of visually uniform pigs is feasible without physical tags, a step toward non-invasive, continuous welfare monitoring in precision livestock farming.

African swine fever (ASF) poses a serious economic threat to European pig production, with no vaccines or treatments available. This study consolidates the scattered economic evidence on ASF costs and biosecurity expenditure and applies a benefit-cost framework to assess whether prevention-focused investment is a sound economic strategy relative to outbreak response. A structured literature search across PubMed/MEDLINE, Web of Science, ScienceDirect, CAB Abstracts and EconLit (January 1990–December 2025) screened 1,247 records, of which 34 were retained to parameterise an illustrative expected-value benefit-cost analysis, with all costs converted to 2025 EUR. For a 122-sow reference farm, preventive biosecurity was estimated at €125,750 infrastructure plus €22,051 annual operating costs (€12.37 per pig), whereas an outbreak imposed farm-level losses of approximately €838,000, exceeding 830% of annual profits (€101,000). European simulations for larger industrialised herds suggested per-herd direct costs of €0.9–1.2 million, while government-funded control reached €28.4 million even for a geographically limited wild boar outbreak with no domestic pig infections, alongside export losses from trade bans. At base-case assumptions (10% outbreak probability, 50% prevention effectiveness), the farm-level benefit-cost ratio was 1.21:1, rising to 2.42:1 at 20% probability; the break-even threshold was 41.3% prevention effectiveness. Because farm-level analysis captures only part of total costs—indirect trade losses may reach 29 times direct control costs—the sector-wide case for prevention is substantially stronger. Despite considerable uncertainty, the evidence consistently favours prevention over reactive approaches, suggesting policy should prioritise biosecurity investment, particularly for vulnerable small-scale producers facing the highest per-unit costs and greatest outbreak risk.

Periparturient and lactating sows need to turn around when expressing important nest-building and piglet-gathering behaviors. To determine the minimum pen width required to allow unhindered turning, 26 Danish crossbred sows were selected in late gestation to represent younger (parity 2-4) and mature (parity 5-6) ages. A test pen measured 2.6m on one side, with the other side being adjustable in intervals of 0.2m (1.2–2.2m). Each sow was encouraged to make three food-motivated turns at each of the 6 pen widths according to a randomized design. Direct observations of the time taken to complete a turn and a subjective score of turning difficulty (range 1-6) were supplemented by automated analysis of video recordings, measuring the maximum curvature of the body during a turn and the proximity to the pen walls at the time of maximum curvature. Turns took longer to complete when pen width was 1.2m and 1.4m. Inability to complete some turns occurred at 1.2m and more turns were scored as hindered when pen width was ≤1.6m. Older, larger sows were more frequently hindered. Body curvature during a turn was greater when pen width was 1.2m, while distance from pen walls at maximum curvature only increased in pen widths of ≥1.8m. In conclusion, the trial indicated that a minimum width of 1.6m should be provided to allow unhindered turning for the full grown late pregnant sows. This should be factored into a suitable overall farrowing pen design for loose sows to avoid negative impacts on sow and piglet welfare.

Background: The objective of this study was to investigate the validity of the Brix refractometer to accurately estimate immunoglobulin G (IgG) concentrations in porcine colostrum, using radial immunodiffusion (RID) as a reference standard. Methods: Eighty-seven composite colostrum samples were collected from sows on a single pig farm between September 2022 and September 2024. Brix measurements were compared with IgG RID test results for each colostrum sample. Results: Brix measurements ranged from 10.80% to 28.05% (mean = 20.13; standard deviation = 3.78). Samples from 40 of the 87 sows (45.98%) were below the colostrum quality threshold of 50 g/L IgG. Pearson's correlation coefficient revealed a strong correlation of 0.81 (p < 0.001) between Brix (%) and RID IgG concentration. Receiver operating characteristic (ROC) curves were constructed to determine the optimal threshold to accurately determine poor colostrum quality (<50 g/L IgG) using the Brix refractometer. At an RID IgG concentration threshold of less than 50 g/L, ROC analysis (Youden index) defined an optimal Brix threshold of 20.05% (sensitivity = 0.75, specificity = 0.75 and area under the curve = 0.76). Limitations: The retrospective design of this study led to some inherent limitations with the quality and quantity of the data collected. Conclusions: The results of this study indicate that the Brix refractometer can be used as an accurate pen-side tool for estimating IgG concentrations in porcine colostrum.

Maternal stress during gestation can alter offspring physiology, behaviour, and immune function. In pigs, such ‘prenatal stress’ is known to increase stress sensitivity, but the potential to automatically detect such sensitivity has remained unexplored. Automatic detection of facial expression has successfully identified differences in pigs dependent on their stress status. This study progresses this work by demonstrating that, for the first time, using a deep learning framework applied to facial analysis, stress-linked phenotypes can be learned from one generation and detected in the next. Using a dataset of over 7000 facial images from 18 gestating sows and 53 of their daughters, we trained and evaluated five state-of-the-art deep learning architectures across six independent daughter cohorts. Attention-based models significantly outperformed CNN-based models, with the Vision Transformer (ViT) model achieving a mean accuracy of 0.78 and an average F1-score of 0.76. Grad-CAM visualisations showed that the ViT consistently attended to biologically relevant facial regions, such as the eyes and snout, whereas CNNs often focused on diffuse or non-informative areas, resulting in reduced low-stress recall and greater batch sensitivity. Models trained on maternal facial images successfully predicted stress responsiveness in daughters from unrelated lineages, indicating that the model captured generalisable facial cues of stress rather than familial resemblance. This approach supports previous work showing that machine vision can detect putatively stress-related alterations to facial expression in pigs. Future application of this approach could offer a scalable, non-invasive tool for early detection of stress in livestock production systems, opening new avenues for welfare-oriented precision livestock management and informed breeding strategies aimed at improving stress resilience.

Stress in pigs poses significant challenges to animal welfare and productivity in modern pig farming, contributing to increased antimicrobial use and the rise of antimicrobial resistance (AMR). This study involves stress classification in pregnant sows by exploring five deep learning models: ConvNeXt, EfficientNet_V2, MobileNet_V3, RegNet, and Vision Transformer (ViT). These models are used for stress detection from facial images, leveraging an expanded dataset. A facial image dataset of sows was collected at Scotland’s Rural College (SRUC) and the images were categorized into primiparous Low-Stressed (LS) and High-Stress (HS) groups based on expert behavioural assessments and cortisol level analysis. The selected deep learning models were then trained on this enriched dataset and their performance was evaluated using cross-validation on unseen data. The Vision Transformer (ViT) model outperformed the others across the dataset of annotated facial images, achieving an average accuracy of 0.75, an F1 score of 0.78 for high-stress detection, and consistent batch-level performance (up to 0.88 F1 score). These findings highlight the efficacy of transformer-based models for automated stress detection in sows, supporting early intervention strategies to enhance welfare, optimize productivity, and mitigate AMR risks in livestock production.

There is pressure to ban the use of farrowing crates. Without a consensus on acceptable alternative systems, commercial progress is delayed and there are risks of fragmented legislation and building of systems that function less well. There is an urgent need for coherent, effective measures to stimulate forward progress. The main technical points of contention are the allowance of temporary crating and confinement time, pen space, enrichment and flooring. We present a multi-criteria review of these issues, designed to be of use to policymakers and other stakeholders. This review synthesizes the European Food Safety Authority (EFSA) recommendations, the latest scientific evidence base and commercial experience. We discuss potential alternatives and recommend actions for the regulation of close-confinement farrowing and lactation systems. We propose a minimum pen footprint of 6.5m2 with linear distances to support unhindered turning by the sow and, when using temporary crating, this should be for no more than 5 days. Flooring type and enrichment provision interact, with regional climatic differences making harmonized legislation about the best options not feasible. The compromises inherent in trying to meet all stakeholder’ needs may result in legislation which seemingly only makes a relatively small difference to the welfare of an individual sow and litter, but a large population of animals will benefit, making the overall increase in pig welfare considerable.

Pregnant (dry) sows are fed a restricted food ration to ensure good health, production and longevity, but this results in behavioural signs of hunger: oral behaviours increase and may be redirected 'unnaturally' towards non-food, prompting welfare concerns. 'Dietary fibre' encompasses a variety of chemical classes, with a variety of physiochemical properties. It reduces abnormal oral behaviours, and fibres that are soluble and fermentable in the hindgut appear to prolong satiety, reduce activity and improve welfare. EU rules require fibre in dry sow diets and access to foraging materials but implementation differs across member states. Diet, feeding system, number of meals and social structure combine to affect welfare of the sow and developing piglets, e.g., through pre-natal stress. Current breeding trends to increasing litter size and associated practices like the use of 'nurse sows' may require a rethink of sow nutrition to safeguard longevity and welfare.

Animal welfare is inherently related to animals’ needs and wants. Needs relate to biological requirements, while wants relate to animals’ desires, which may go beyond what is required to maintain health. The Five Freedoms, which underpin much of the guidance and legal frameworks regarding protection of farm animal welfare, are closely related to animals’ needs. Yet despite their acceptance as the basis for minimal welfare standards, it can be argued that many of the freedoms, and thus animals’ needs, are not actually met for pigs kept under commercially intensive conditions. This is particularly the case regarding the ‘Freedom to express normal behaviour’. In many cases, normal behaviour is not merely a desire but innate and needs to be performed in order to remain physically or psychologically healthy or to produce healthy offspring. This chapter aims to reflect on pigs’ needs and wants and how these are met in current pig husbandry systems. We also propose a matrix of needs and wants as guidance for moving towards better standards.

Acceptable animal welfare is an integral part of sustainability. Selective breeding for improved animal welfare can benefit the economic and environmental aspects of pig farming, as well as being of direct benefit to the animal itself. Several traits have major welfare consequences but have proved difficult to improve through management change alone. Here we consider how past selection for productivity has affected welfare and give three examples of the state of the art in selective breeding aimed at improving welfare traits in their own right. Selection for complex welfare-relevant traits poses practical, economic, and ethical challenges. Current and emerging innovations will significantly reduce the economic and practical barriers to breeding and allow efficient selection for traits that previously have been too expensive to record. Examples will be given of the new phenotyping techniques and genetic methodologies that are expanding the range of welfare traits that selection pressure can be exerted upon.

Abstract Background Pressure to abolish farrowing crates is increasing, and producers are faced with decisions about which alternative system to adopt. For sow welfare, well designed free farrowing systems without close confinement are considered optimal but producers have concerns about increased piglet mortality, particularly crushing by the sow. Reporting accurate performance figures from commercial farms newly operating such systems could inform the transition process. This study investigated performance on three commercial farms operating four different zero-confinement systems, three of which were newly installed. A total of 3212 litters from 2920 sows were followed from farrowing to weaning over a three-year period with key performance indicators (KPIs) recorded. Mixed Models (LMMs, GLMMs) determined the influence of different factors (e.g. farrowing system, sow parity, management aspects) and litter characteristics on performance, including levels and causes of piglet mortality. Results Piglet mortality was significantly influenced by farm/system. Live-born mortality ranged from 10.3 to 20.6% with stillbirths ranging from 2.5 to 5.9%. A larger litter size and higher parity resulted in higher levels of mortality regardless of system. In all systems, crushing was the main cause of piglet mortality (59%), but 31% of sows did not crush any piglets, whilst 26% crushed only one piglet and the remaining sows (43%) crushed two or more piglets. System significantly influenced crushing as a percentage of all deaths, with the system with the smallest spatial footprint (m2) compared to the other systems, recording the highest levels of crushing. Time from the start of the study influenced mortality, with significant reductions in crushing mortality (by ~ 4%) over the course of the three-year study. There was a highly significant effect of length of time (days) between moving sows into the farrowing accommodation and sows farrowing on piglet mortality (P < 0.001). The less time between sows moving in and farrowing, the higher the levels of piglet mortality, with ~ 3% increase in total mortality every five days. System effects were highly significant after adjusting for parity, litter size, and days pre-farrowing. Conclusion These results from commercial farms demonstrate that even sows that have not been specifically selected for free farrowing are able, in many cases, to perform well in these zero-confinement systems, but that a period of adaptation is to be expected for overall farm performance. There are performance differences between the farms/systems which can be attributed to individual farm/system characteristics (e.g. pen design and management, staff expertise, pig genotypes, etc.). Higher parity sows and those producing very large litters provide a greater challenge to piglet mortality in these free farrowing systems (just as they do in crate systems). Management significantly influences performance, and ensuring sows have plenty of time to acclimatise between moving in to farrowing accommodation and giving birth is a critical aspect of improving piglet survival.

Piglet mortality is a problem with complex aetiology, predisposed by the natural biology of a polytocous species which produces multiple offspring allocated disproportionate resources both in the pre- and post-natal environment. Intensive sibling rivalry to acquire limited uterine resources prenatally, and then limited nutrients postnatally, leads to a high risk of mortality for the weakest among the offspring. In the domestic pig these risks have been exacerbated by intensive selection for economic traits such as prolificacy and leanness. Therefore a certain amount of mortality would seem expected. The acceptability of death can be a subject of much ethical debate. However, even if death itself is not considered a welfare issue, the manner of dying is, particularly when it involves pain and/or suffering. Thus, whilst the life history strategy of the pig encompasses mortality risk, at what point does the inevitable become unacceptable for animal welfare?

Most sows worldwide farrow while restrained in a farrowing crate. This poses a welfare dilemma. The restriction of sow movement interferes with the performance of species-specific behaviours such as nest-building, orientation, exploration, and communication with the piglets and leads to increased physiological stress. However, allowing the sow more freedom to perform motivated behaviours often results in increased crushing of piglets and hence a piglet welfare problem. Research efforts have led to successful development of loose farrowing systems delivering high performance and high welfare. Key to success is allowing species-typical behaviours that contribute to biological fitness, which encompasses important performance parameters such as number of offspring produced, offspring viability, and maternal rearing ability. Performance and welfare can improve by optimising system design (e.g. housing), human inputs (e.g. management), and animal inputs (e.g. genetic selection strategies). Continuing genetic selection pressure for sow leanness and prolificacy has, however, significant impacts on their welfare. A more balanced equation, where the cost of maternal investment is balanced against the increase in the number of piglets born and weaned, is needed as a basis for breeding programs.

Gastrointestinal (GI) parasitism is a health and production concern in sheep, yet its impact on animal welfare remains unclear. The impact of subclinical infections is especially ambiguous as GI parasitism often remains undiagnosed until clinical signs such as diarrhoea are evident. This study applied quantitative and qualitative methods to examine the effects of subclinical Teladorsagia circumcincta infection on the behaviour and welfare of 96 Suffolk-cross lambs (24 pens of 4 lambs) weaned at 10 weeks old. The hypothesis that parasitism causes negative affective states was tested. Lambs were divided into three groups at the pen level: ad-lib fed control (AC), restricted-fed control (RC), and ad-lib fed parasitised (AP). Parasitised lambs (AP) were dosed three times weekly with 7000 third stage T. circumcincta larvae (L3) from 16 weeks of age. Lambs in the RC group were pair fed to match AP feed intake to separate the effects of infection-induced anorexia from the potential direct impacts of infection. From 7 days pre-infection to 23 days post-infection, scan and behaviour samples were taken from video recordings to quantitatively monitor behaviour, and animal-based measures such as faecal soiling score (FSS) were recorded as welfare indicators. Lying, standing, eating, play and social behaviour were monitored. Qualitative behaviour assessment (QBA) was conducted weekly using the AWIN (2015) protocol to gain insight into the lambs’ affective states over the onset of infection. Parasitised lambs were more likely to stand inactive than AC lambs as the infection progressed (P=0.006). They were also less likely to display eating behaviour in the third daily scan sample than RC lambs (P<0.001). Principal Component Analysis of the QBA data revealed that the first dimension (PC1) described arousal levels, the second (PC2) described the valence of the animals’ affective states, and the third (PC3) described fearfulness and aggression levels. Parasitised lambs (est=10.64,SE=0.33) scored higher than RC lambs (est=9.42, SE=0.33) on PC3, the fearfulness dimension (P=0.030). There were no differences between fearfulness scores of AC and AP lambs or RC lambs and treatment group had no significant impact on the distribution of scores on PC1 or PC2. These findings demonstrate that subclinical GI parasitism negatively impacts lamb welfare not only in the health domain but in the behaviour and mental domains as well. This has implications for welfare assessments and early disease detection in lambs. Future research could explore remote monitoring of the indicators of parasitism identified in this study.

The importance of maternal care in commercial pig production is largely ignored. The sow has little possibility to interact with her piglets, and piglets are often subjected to early weaning or artificial rearing. This study aimed to investigate aspects of physiological and behavioural maternal provisioning that contribute to offspring outcomes. We hypothesised that better maternal care and nutritional provisioning would relate positively to piglet immunity, growth and behaviour. Nineteen sows and their litters were studied in free-farrowing pens. Oxytocin and tumour necrosis factor-α in colostrum/milk and salivary cortisol were sampled from sows throughout lactation. Sows were assessed for dominance rank, response to handling, maternal defensiveness, suckling initiation and termination, posture and sow-piglet contact. Piglets were weighed, measured for body mass index (BMI) and sampled for blood (Immunoglobulin G; at birth). After weaning, they experienced a human approach test (HAT) and novel object test. Correlations were explored between individual sow characteristics, individual piglet outcomes, and between sow characteristics and piglet outcomes averaged by litter. Significant correlations between sow and piglet factors were analysed at the litter level in mixed models with piglet outcomes as response variables and sow characteristics as predictor variables, while accounting for sow parity, litter size and batch. Litters grew faster when their sow had lower cortisol values (P = 0.03), while sows with lower cortisol levels had more successful suckling bouts and engaged in greater amounts of sow-piglet contact. Litters had a lower BMI at weaning when the sow had a higher milk fat percentage at d3. Litters of the most dominant sows took longer to approach the human in the HAT, while litters of sows with higher cortisol at d0 took longer to approach the novel object when assessed on correlations (r = 0.82, P < 0.001) but not when the model accounted for parity and litter size (P = 0.35). Only some of the measured nutritive and non-nutritive sow factors influenced litter performance and behaviour, with parity and litter size also playing a role. Given the continued increase in litter size, but also the interest in loose-housed lactation pens for sows, further research on sows' maternal investment and how it can be optimised is warranted.

Alternatives to carbon dioxide (CO2) stunning for the commercial slaughter of pigs are urgently needed because there is robust evidence that exposing pigs to hypercapnic environments is associated with pain, fear, and distress. Hypobaric hypoxia (via gradual decompression, also known as Low Atmospheric Pressure Stunning or LAPS) has been validated in poultry as a humane option, but its potential to improve the welfare of pigs at slaughter is unknown. We investigated the potential of hypobaric hypoxia to reliably elicit a non-recovery state in anesthetized weaner-grower pigs within a commercially viable timeframe. We determined the effect of candidate decompression rates (40, 60, 80, 100 ms(-1), at two cycle durations 480 s and 720 s) on a range of physiological and reflexive behavioral indicators of hypoxia and death. We found that the decompression rates tested caused a 100% death rate. As expected, the decompression rate had overarching effects on behavioral and physiological markers of hypoxia and death, with faster decompression rates resulting in shorter latencies to cardiac arrest and cessation of breathing. We observed a higher proportion of pigs displaying repeated and prolonged whole-body movements (likely indicative of convulsive activity) at higher frequencies when we applied the slowest decompression rate (40 ms(-1)) compared to all other rates. Since these responses may impact the carcass and meat quality, the slower rate of decompression (40 ms(-1)) should be excluded as a candidate decompression rate. Furthermore, given the marginal effects of decompression rate on physiological indicators of death and reflexive behavioral parameters, we also recommend that the fastest rate tested (100 ms(-1)) is excluded in further study on conscious pigs (to prevent conscious animals from being exposed to unnecessary faster decompression rates which may compromise animal welfare). This work represents a necessary proof of principle step and confirms the potential of gradual decompression for stunning purposes in pigs. Importantly, however, the data presented provide no information on the welfare outcomes associated with decompression in conscious pigs. Subsequent work should focus on the comprehensive welfare assessment of intermediate decompression rates to determine the potential of hypobaric hypoxia to provide a humane stunning method for pigs.

There are animal welfare concerns about the continued use of permanent crating systems for farrowing and lactating sows, which is the most prevalent maternity system in global pig production. Greater societal attention in recent years has culminated in changes (or proposed changes) to regulations as well as market-driven initiatives to move away from crated systems. Transitioning from farrowing crates to systems that allow the sow greater freedom of movement and behavioral expression requires a number of key decisions, with various trade-offs apparent when trying to balance the needs of different stakeholders. This review discusses these decisions based on common questions asked by farmers, policy makers and other stakeholders when deciding on a new system to build/approve. Based on the latest scientific evidence and practical insight, decisions such as: whether to retrofit an existing barn or build a new one, what spatial dimensions are necessary per sow place, whether to adopt free farrowing or temporary crating, how to provide substrate/enrichment and be hygienic and environmentally friendly, and how to optimize the human inputs and transition between systems are considered. The aim of this paper is to provide a roadmap for those interested in uptake of higher welfare systems and practices, as well as to highlight areas requiring further optimization and research.