Current methods used to assess pain in farm animals focus on changes in behavior, physiology and production/bodily functions. For example, a reduction in play behavior was noted after castration in lambs [55
], and after disbudding in calves [74
]. Devant et al. [53
], noted feed intake declined after castration, as did lying behavior and overall activity levels. Bonastre et al. [91
], assessed the physiological responses of piglets to castration, finding surface skin temperatures to drop immediately after castration independent of any anesthesia or analgesia given. They also found glucose concentration to increase in response to castration, and cortisol concentrations to differ in response to castration with or without anesthesia and analgesia. There are also a number of studies demonstrating the significant negative impact of lameness in dairy cattle on milk yield [92
]. It must be remembered that there are a number of other possible factors that can affect production, and not just pain, making productivity an unreliable measure of current affective state.
There are numerous limitations to the collection and interpretation of these measures. Many of these measures act more as a long-term indicator of the impact of pain, meaning that they cannot provide a true reflection of how an animal may be feeling at that moment. Collection of some of the physiological measurements require the animal to be restrained, compromising the ability to interpret changes as painful as opposed to stressful. Other pain assessment systems have been developed in response to this, and they rely on monitoring short-term behaviors such as vocalizations [96
], escape attempts [97
], lameness [94
], or posture changes [57
] where animals try to adopt a position that limits the pain experienced. Many of these indicators are not pain specific and can be seen in response to other affective states such as fear or stress, or even positive states such as joy.
For any pain assessment tool to be of value, it must be valid, reliable and feasible [52
]. It must enable recognition, assessment and evaluation of any pain experienced in a sensitive and specific manner. The pain should change in response to analgesics in a dose dependent manner [100
] to demonstrate that pain is present and that it is ameliorated with pain relief; this can be particularly problematic in farm animal’s due to the lack of licensed pain relief and a lack of literature of the effectiveness of the available drugs. Any new assessment tool must be tested against an already existing, validated tool; in humans, this is verbal confirmation of their experience. Animals are unable to verbally confirm their experience with humans, and therefore validation can become difficult, particularly with new diseases or conditions. A measurement must also be feasible for use on farm being non-invasive, easy to use without the need for any specialist equipment, and needing only minimal training.
Pain is highly complex and multifaceted, so any measure must be able to consider the intensity, frequency, and duration of pain experienced [101
]. This can be particularly difficult due to a number of other factors affecting how an individual may perceive the pain experienced, including any previous experience of pain they may have had, and when this experience occurred [57
]. An individual’s personality may also affect their perception of pain [104
], along with their sex [54
5.1. The Use Facial Expression to Identify and Evaluate Pain in Farm Animals
Facial expression has been used extensively in human research and medicine as a way to assess pain in non-verbal patients. Facial expression is the measurement of changes in the face or groups of muscles, known as “action units” to an emotional stimulus. Facial expression is considered to be an honest signal of the affective state and intensity of the pain [105
]; it becomes increasingly difficult to “hide” the expression of pain in the face [106
], and faked pain expression is easily identified [107
]. Aversive feelings are expressed differently within the face [109
], and there is evidence for both the affective and sensory component of pain to be expressed within the face through different action unit movements [110
]. The expression, and sensitivity of the expression of pain, is also conserved across development in humans [111
It has only been in the last decade that facial expression has been developed for use in animals as a pain assessment tool (mice [113
], rats [114
], rabbits [73
], horses [71
]), including some farm species (sheep [32
], lambs [116
] and piglets [117
]). Across the different species, there are similar facial movements and action units expressed in the presence of pain, demonstrating an evolutionary stability in pain expression across mammalian species [118
]. Langford et al. [113
], were also able to demonstrate that facial expression of pain could be separated from the sensory (abdominal writhing) expression of pain in mice, albeit in a small sample size (n
Facial expression allows for immediate and spontaneous identification and assessment of pain in animals which is vital for effective pain management; this is a distinct advantage over a number of other pain assessment tools which rely on retrospective assessments. In addition, observers are naturally drawn towards the facial area [119
] making facial expression as a pain assessment tool, particularly good. Minimal training is required, and once the “grimace scale” has been learnt the technique is easy to use, without the need for specialist equipment. The scales have been shown to be valid and have high inter-observer reliability scores, ranging from 85% [71
] to 97% [117
]. High reliability should mean that veterinarians, paraprofessionals such as veterinary nurses, and animal carers assess pain in a systematic and consistent manner, ensuring that the animal receives consistent care and management of its pain [120
]. This also means that both vets and farmers should be able to monitor and assess pain in their animals the same way, allowing for a more agreed take on the pain experienced enabling appropriate treatment.
It is likely that facial expression is an involuntary response to pain experienced by an animal [113
], leading to a higher sensitivity of pain assessment compared with other assessment tools. The use of the scales also provides a more accurate assessment of the pain than the more subjective global assessment of pain [32
]; this is likely due to the removal of a need to make a decision about the animal. The scales assess each part of the face and observers only need to decide on a score for each specific part, rather than the animal as a whole. Facial expression also allows for the assessment of the temporal nature of pain; whether there is a high degree of fluctuation or if there is constant pain. Measuring the temporal nature will allow for observers to have a better understanding of the frequency and duration of the pain, enabling the development of a better pain management strategy. It also allows observers to monitor the effects of any pain relief that has been provided and to adjust it accordingly. Prompt recognition of pain allows for prompt management of the pain, improving animal welfare.
Some of these scales have also provided an intervention threshold [32
], again removing the need for observers to make a decision about the need to give pain relief or not. Having a scale that can objectively identify and assess pain and provide guidance on when to give pain relief will improve the chances of that animal receiving treatment. Pain relief should change the facial expression of animals in pain [113
] demonstrating if it is working or not. This can be particularly useful when giving pain relief to animals that do not have a licensed drug, or where there is limited research into the effectiveness of the drug.
Individuals experience pain differently. Facial expression gives the ability to assess each individual as such. The facial expression scales provide a more feasible long-term assessment of pain, capturing any fluctuations in pain experienced. Keeping records of an animal’s facial expression over time will allow for a better understanding of how that individual may be coping, and whether more pain relief is required, or a different drug needed. Facial expression provides a valid and reliable technique to assess pain in farm animals; however, many of these scales are in their early stages of development and require feasibility testing. In addition, there are some farm animal species missing a validated scale, such as cattle. The technique is more likely to be utilized as a pain assessment tool on farm in real-time once there is data showing its feasibility.
Automation of Facial Expression Detection in Sheep
Manual scoring of the facial expression of any animal has the potential to be biased by other knowledge about the subject such as the presence of disease, lameness, or other behaviors and postures which may provide conflicting information. Although the technique is not as time consuming as other behavioral, physiological or production related measures, it still at present requires time for the farmer or veterinarian to observe the animal. The automation of the analysis of facial expression would improve efficiency as it would not require someone to be present to assess the animals, and it would ensure consistency in the estimation of pain [121
] by removing the subjectivity of the assessment. Automation would also mean that farmers would not need to spend their limited time learning individual facial expressions; the automated system would do this. The author currently collaborates with a team in the Computer Laboratory, University of Cambridge who are developing an automated system for the detection and analysis of facial expression in sheep using data collected by the author. The system is currently in the early stages of development and utilizes technology developed for human emotion detection. It can currently detect individual faces of sheep, localize facial landmarks, normalize and then extract facial features [122
]. Currently the approach can successfully detect action units and assess pain level in sheep using forward facing pictures (frontal), but is still not yet capable of assessing animals via portrait view (see Lu et al. [121
] and Mahmoud et al. [122
] for full details of the working models). The data set available is currently limited, but with more data, the automated approach of assessing pain in sheep will be viable to use on-farm in real-time. It is envisaged that multiple cameras and electronic identification tag readers will be placed at key feeding, water, and resting stations used by sheep around housing or fields. The tag readers will link the identification of the sheep and the image taken. Cameras will monitor the facial expression of sheep visiting these stations, learning the individual appearance over time and identifying any sheep that shows indications of abnormal feature position as described in the Sheep Pain Facial Expression Scale (SPFES) [32
], alerting the farmer to any animal that registers a facial expression score at or above the pain threshold detailed in McLennan et al. [32
]. It also has the potential to be applied to different mammals and across different situations e.g., on farm or at market [121
The use of technology to optimize production and management of each individual animal is becoming key in efficient and effective farming. Many technologies are currently available for use on farm, learning about the individual [123
], and getting to know what is normal for each individual, rather than a general herd assessment. An automated detection of the facial expression of a sheep would improve the chances of detecting a change in the sheep’s normal facial expression suggesting the presence of pain. This would allow for intervention and provision of treatment to that individual earlier than what might have otherwise occurred when there was a need to assess in person. Early detection of any health problem will ensure that the animal can get back on track to full health as soon as possible, reducing the impact on welfare and productivity.