A team of researchers led by Dr. Vivek Kumar at the Jackson Laboratory (JAX) has developed a groundbreaking method to weigh mice without causing them stress.
This new technique uses computer vision, allowing scientists to accurately and continuously measure the body weight of mice without having to handle them.
The goal is to improve the quality and reliability of biomedical research that relies on these tiny animals.
In both human health and research, body weight is an important measure that can give clues about overall health and potential health problems.
Mice are commonly used in preclinical studies because their biology is similar to humans in many ways.
However, traditional methods of weighing mice require taking them out of their cages and placing them on a scale.
This process can be stressful for the mice, which can lead to inaccurate data and affect the results of experiments.
“We realized that there was a need for a better way to accurately and noninvasively measure the weight of animals over time,” said Dr. Kumar.
“The old method stresses the mice and limits how often we can measure them, which can weaken the accuracy of our experimental results.”
To solve this problem, Dr. Kumar and his team, including first author Malachy Guzman, turned to computer vision technology. Guzman, who started this project as a summer intern at JAX and continued it during the school year, worked alongside other team members like Brian Geuther and Dr. Gautam Sabnis.
Together, they analyzed one of the largest video datasets of mice ever collected. Previously, this dataset had been used to study behaviors like grooming and walking posture.
The researchers developed a computer vision method that calculates the body weight of mice with an accuracy of less than 5% error.
Their findings were published in the August 7 issue of the scientific journal Patterns.
This new technique is a valuable resource for scientists because it allows them to get more precise and reliable data in their studies.
One of the biggest challenges the team faced was that mice are very active and constantly change their posture and shape. Unlike larger animals in industrial farming, mice move around a lot, making it harder to measure their weight accurately. The team also worked with 62 different strains of mice, each with different sizes, behaviors, and fur colors. To overcome these challenges, the researchers used a variety of visual metrics, machine learning tools, and statistical models.
“Although only 0.6% of the video pixels belonged to each mouse, we were able to use computer vision to predict their body weight,” Guzman explained.
By training their models with diverse strains of mice, the researchers ensured that their method could handle the wide range of appearances and sizes seen in lab mice.
This new technique has several advantages. It allows researchers to detect small but important changes in body weight over several days, which could be crucial for studies involving drugs or genetic experiments.
It also has potential as a diagnostic tool for general health monitoring and could be adapted for other experimental environments and even different organisms in the future.
