AI assessment indicates stress levels in farmed Amazonian fish

In Brazil, a group led by researchers from São Paulo State University (UNESP) in Jaboticabal, in collaboration with the Brazilian Agricultural Research Corporation (EMBRAPA), has developed an artificial intelligence (AI) tool to assess the stress levels of tambaqui (Colossoma macropomum), the most widely produced native fish in Brazil. The study is published in the journal Aquaculture.

The results could impact both animal welfare and the selection of specimens that are more tolerant of the farming environment. Tambaqui is an Amazonian species primarily farmed in the northern states. Brazil is the world’s largest producer of the species, supplying 110,000 tons in 2022.

“First, we found that in stressful conditions, that is, in a more confined environment than normal, the fish became darker. Then, we saw that the addition of a stress-related hormone also altered the coloration of the scales. So we trained software with more than 3,000 images to arrive at a stress threshold that could guide fish farmers and genetic selection programs, as we saw that this is a heritable trait,” explains Diogo Hashimoto, a professor at the UNESP Aquaculture Center (CAUNESP) who coordinated the study.

Celma Lemos, a doctoral student at the institution, is the first author of the study.

To develop the tool, the researchers photographed 3,780 tambaqui from two populations: one from CAUNESP (1,280 individuals) and the other from EMBRAPA Fisheries and Aquaculture in Palmas, in the state of Tocantins (2,500 individuals), with the collaboration of the team coordinated by researcher Luciana Shiotsuki.

Next, the lower half of the body was marked on each image for evaluation by the software. Contrast with the coloration of the upper part is common in fish and is probably an attribute of natural selection that has resulted in a kind of camouflage. This phenomenon, known as countershading, can be observed in sharks, for example, which have a lighter belly than their backs.

The researchers then trained a deep learning model to determine a threshold based on the ratio of black to white pixels in an image to indicate the degree of stress in the tambaqui.

Since the Tocantins specimens had been marked according to their ancestry, they were able to determine the extent to which the trait can be passed on.

“We estimate that stress tolerance is a moderately to highly heritable trait. This is reflected in weight gain and disease tolerance, paving the way for generations with increasing well-being in a farming environment,” says Hashimoto.

Physiological mechanisms

Several fish species exhibit a change in coloration under stress, but this trait had not yet been proven in tambaqui. In species that become darker under stress, such as tilapia (Oreochromis niloticus), a stress-related hormone promotes the expansion of melanophores, which are cells that appear as small black spots to the naked eye.

To demonstrate this effect in tambaqui, the researchers collected scales from six individuals of the Jaboticabal population and immersed them in two solutions. One solution contained a neutral substance and α-MSH (a version of the melanocyte-stimulating hormone), and the other contained only the neutral substance. After 30 minutes, the researchers observed that the scales bathed in the hormone solution were darker, with expanded melanocytes.

In another experiment, six tambaqui were removed from their normal 200-square-meter breeding tanks. They were photographed and then divided into three much smaller, round reservoirs. Each reservoir had a capacity of 2,000 liters and measured 85 centimeters in height and 1.66 meters in diameter. After ten days, the fish were photographed again, revealing a difference in coloration that confirmed the species becomes darker when stressed.

“The AI tool can be used to monitor the stress of farmed fish at a time when animal welfare is increasingly demanded. By simply evaluating photos of the animals, it would be possible to obtain this measurement and improve practices when necessary, such as reducing the number of individuals per tank, for example,” says the researcher.