Researchers develop stress-free methodology to weigh mice utilizing pc imaginative and prescient

A analysis workforce led by Jackson Laboratory (JAX) Affiliate Professor Vivek Kumar, Ph.D., has developed a non-intrusive methodology to precisely and constantly measure mouse physique mass utilizing pc imaginative and prescient. The method goals to cut back the stress related to conventional weighing methods, enhancing the standard and reproducibility of biomedical analysis involving mice.

In each human well being and biomedical analysis, physique mass is a essential metric, typically serving as an indicator of total well being and a predictor of potential well being points. For researchers working with mice—the commonest topics in preclinical research—measuring physique mass has historically concerned eradicating the animals from their cages and inserting them on a scale.

This course of will be demanding for the mice, introducing variables that may have an effect on the result of experiments. Furthermore, these measurements are sometimes taken solely as soon as each few days, additional complicating the accuracy and reproducibility of knowledge.

“We acknowledged the necessity for a greater methodology to precisely and noninvasively measure animal mass over time,” mentioned Kumar. “The standard method not solely stresses the mice but additionally limits the frequency and reliability of measurements, which may weaken the validity of experimental outcomes.”

To handle this problem, Kumar and his workforce of computational scientists and software program engineers, together with first writer Malachy Guzman, in addition to Brian Geuther, M.S. and Gautam Sabnis, Ph.D., turned to pc imaginative and prescient know-how. Guzman, a rising senior at Carleton School, was a JAX summer time pupil and continued the venture as a tutorial yr intern with the Kumar Lab.

By analyzing one of many largest mouse video datasets utilized by Kumar beforehand to additionally assess grooming conduct and gait posture, they developed a way to calculate physique mass with lower than 5% error. The findings, revealed within the Aug 7 superior on-line challenge of Patterns, provide a brand new useful resource for researchers that guarantees to boost the standard of a variety of preclinical research.

The analysis workforce confronted a number of challenges in growing this methodology. Not like the comparatively static topics utilized in industrial farming for physique mass measurement, mice are extremely energetic and versatile, ceaselessly altering posture and form, very like a deformable object.

Moreover, the workforce labored with 62 completely different mouse strains, ranging in mass from 13 grams to 45 grams, every with distinctive sizes, behaviors, and coat colours, necessitating using a number of visible metrics, machine studying instruments, and statistical modeling to realize the specified degree of accuracy.

“Within the video information, solely 0.6% of the pixels belonged to every mouse, however we had been in a position to apply pc imaginative and prescient strategies to foretell the physique mass of particular person mice,” Guzman defined.

“By coaching our fashions with genetically various mouse strains, we ensured that they may deal with the variable visible and dimension distributions generally seen in laboratory settings. Finally, our statistical fashions to foretell mass can be utilized to hold out genetic and pharmacological experiments.”

This new methodology affords a number of key benefits for researchers. It allows the detection of small however vital modifications in physique mass over a number of days, which could possibly be essential for research involving drug or genetic manipulations.

Moreover, the strategy has the potential to function a diagnostic software for normal well being monitoring and will be tailored to completely different experimental environments and different organisms sooner or later.