The Pic’Let project, which started in January 2020, is the result of a collaboration between Alliance R&D, which is composed of several French pig breeding companies including Choice and IFIP (the French national pig institute) with Neotec-Vision (a partner specialising in digital imagery).
What is called the maturity phenotype of the piglet?
For several years now, genetic progress on the number of born alive has greatly improved in the Large White and Landrace French reference maternal lines (+ 1.9 born alive between 2014 and 2018) and is still improving. At the same time, the number of weaned is also increasing but at a slower rate. Thus, the mortality of piglets before weaning is increasing.
Reducing mortality in the first few days of a piglet’s life remains a considerable challenge from both an animal welfare and an economic point of view.
Research has shown that a piglet weighing less than 1.2 kg at birth is less likely to survive. So, birth weight may be one of the leads for selection against pre-weaning mortality. For this reason, “piglet birth weight” is one of the traits that has been included in the selection objectives of Choice in the M3 Large White and M6 Landrace maternal lines for several years.
However, the birth weight of the piglet alone does not explain everything, because there are two types of small piglets. A piglet with a low birth weight can be normally developed or it may have intrauterine growth retardation. This intrauterine growth retardation is expressed by a priority of brain development over other organs, which results in lower survival during lactation and lower production performance if the piglet survives. The different degrees of brain development relative to other organs is called the piglet maturity phenotype. There are 3 different phenotypes:
- Normal maturity phenotype: defines a well-developed piglet (A).
- Light maturity phenotype: defines a piglet with a small intrauterine growth retardation, which express itself by a slightly rounded skull, a smaller snout, slightly bulging eyes and thin ears (B).
- Severe maturity phenotype: defines a piglet with severe intrauterine growth retardation. The piglet has a very rounded skull, a very small snout, bulging eyes and thin ears that stick to the neck (C).
One way that this phenotype can be used in a genetic selection program is by calculating the trait “Proportion of immature piglets in the sow” which will be analysed as a trait of the dam. For each farrowing, the number of light and severely immature piglets is counted and a proportion to the total number of piglets is calculated:
A 2018 study (Matheson et al.) shows that this trait is heritable at 0.19 and positively correlated with litter size at 0.23, meaning that the bigger the litter, the higher the proportion of immature piglets. Since this is unfavourable, it may represent a new selection tool to improve litter quality and reduce preweaning mortality.
Prototype & protocol in use at Choice farm
The goal of the Pic’Let project is to develop a tool capable of measuring this maturity phenotype in an automatic and objective way on selection farms.
This tool will take standardised photos of the piglets’ head while the animal is held immobile in a special device. The device is composed with a gutter where the piglet can rest its head and a sliding system, therefore adaptable, which keeps the piglet’s hindquarters higher than its head to prevent it from moving. This process is painless, quick, and replaces the subjective scoring done by the farrowing attendant.
The photos are taken at the time of weighing the piglet in the first 24 hours of life. The prototype, developed by Neotec-Vision, uses a deep learning algorithm to analyse the photos according to the previously mentioned characteristics that mark the head of the immature piglet.
The deep learning method is based on a learning phase of the algorithm. During this initial phase, the algorithm will be trained to recognise an immature piglet using a bank of images carefully chosen and annotated according to the maturity phenotype of the piglets. The algorithm will slice the image into small characteristic areas to build a learning base.
The more images provided to the algorithm, the better the algorithm will be able to distinguish between the different maturity phenotypes over time. Next comes the test phase, where random pictures of piglets are presented to the algorithm. This phase is crucial in determining how well the algorithm works.
For each population measured (M6 Landrace and M3 Large White), the large majority of piglets have a normal maturity phenotype. All three maturity phenotypes are found among piglets weighing less than 1 kg at farrowing. The heavier a piglet is at birth, the less likely it is to have an immature phenotype.
Today, the algorithm can detect an immature piglet with a sensitivity of 89% and a specificity of 74%, meaning that the algorithm will detect an immature piglet in 89% of cases and a mature piglet in 74% of cases. These first results are encouraging but it is important to confirm the observed trends by increasing the number of images of piglets at birth.
This prototype project is opening the way for an automated and objective way to collect the phenotype without additional work for the farrowing staff. By allowing continuous collection and recording piglet maturity at farrowing, and understanding its role in piglet survival, we have the possibility of using the trait to increase the profitability of Choice clients around the world.
References
André L., Manceau J., Gauthier V., Liaubet L., Bonnet A., Monziols M., Brenaut P., Développement d’un outil d’enregistrement automatique de la maturité du porcelet. Journées Recherche Porcine (2021), 53, p. 49-50.
Brenaut P., Outil d’enregistrement automatique de la maturité du porcelet (2021). [online]. In : IFIP, France. Available on : https://ifip.asso.fr/outil-denregistrement-automatique-de-la-maturite-du-porcelet/
Matheson S.M., Wailing G.A., Edwards S.A., Genetic selection against intrauterine growth retardation in piglets : a problem at the piglet level with a solution at the sow level (2018). Genet. Sel. Evol., 50, p. 46.