The media is abuzz with the subject of chronic pain and the corresponding rise in prescription pain killer abuse. The issue affects millions of people around the world, exacting a tremendous cost on health care costs and lost productivity, and imposing great emotional and financial tolls on those who suffer it. Now research using artificial intelligence to diagnose pain in sheep, may yield more data on our own forms of pain.
Human beings are not the only ones who battle pain daily – so, too, do animals, who often fail to express pain in the way we would expect.
Sheep, for instance, often fall prey to infections and disease, yet to the untrained human eye, everything may seem normal. Just a few of the painful conditions they suffer in silence include polyarthritis (which affects the leg joints), foot rot (a very painful condition causing the foot to rot away), mastitis (an inflammation of udders in ewes resulting from infection or injury) and uterine infections. In the case of foot rot, for instance, a sheep will not show the characteristic symptoms (which include lameness) until the disease is quite advanced. This results in unnecessary pain for affected sheep, but also in longer treatments that may not produce the desired outcome if the disease has reached very advanced stages.
Related articles
Artificial intelligence could be a major lifesaver for sheep and other animals undergoing unnecessary pain because of the difficulty of detection, thanks to the work of computer scientists at the University of Cambridge. Their research, presented at a conference in Washington, showed that the well being of sheep (and other animals) could greatly be improved by the early diagnosis and treatment of painful conditions via AI.
The new system detects different parts of a sheep’s face, comparing the face with a standardized measuring tool created by veterinary scientists to diagnose pain. In 2016, another University of Cambridge scientist, Dr. Krista McLennan had developed a useful pain level test called the Sheep Pain Facial Expression Scale (SPFES). The latter enabled sheep farmers to recognize pain by analyzing the facial expression of sheep, yet training people to use the scale properly was time-consuming, and the reliance on individual perceptions meant that accuracy could suffer from case to case.
The newly developed system relied on the SPFES to come up with an AI system which relies on machine learning techniques to provide pain estimates in sheep.
Previously, AI had been used to analyze expressions in humans, but this is the first time the technology has been tested on animals. Lead researcher, Dr. Peter Robinson, noted: “A lot of the earlier work on the faces of animals was actually done by Darwin, who argued that all humans and many animals show emotion through remarkably similar behaviours, so we thought there would likely be the crossover between animals and our work in human faces.”
The SPFES defines five major changes which occur in sheep when they are in pain: their eyes become narrower, their cheek muscles tighten, their ears flop forwards, their lips are stretched down and back, and their nostrils (which are normally U-shaped) take on a V-shape.
The SPFES then ranks the severity of pain depending on the extent of each changed characteristic. Interestingly, the researchers noted that some of these facial changes are also observed in human beings in pain – in particular, the muscles in our cheeks tend to tighten and our eyes become narrower.
The scientists used around 500 photographs of sheep receiving treatment for pain-related conditions, to train the AI model by labelling the different parts of the faces in each photograph, ranking pain according to the SPFES.
Results showed that the AI model was able to estimate pain levels with a remarkable 80 per cent accuracy. The next step for researchers is to train the model to recognize faces through moving images, and to train it to identify the sheep’s faces even when they are not directly facing the camera.
After all, in real life, sheep do not ‘pose’ for cameras and in order to be useful, any AI system would have to be able to predict disease by following animals in normal movement.
Once the system was refined, it could be a very useful tool for farmers, who could take the affected sheep to receive a diagnosis and early treatment from their veterinarian.