As you know, mice and rats are model animals and are often used by scientists in a variety of studies. Moreover, the rodents are very socialized and sociable.
But is it possible to find out what mice and rats "say" to each other during experiments? This question remained unanswered for a long time. The fact is that many characteristic vocalizations of rodents cannot be discerned by human hearing (they are emitted in the ultrasonic range). And existing computer programs for detecting such sounds still require significant improvements. In particular, they cannot filter out extraneous noise and are slow to analyze data, relying on "inflexible" algorithms.
A new program called DeepSqueak (literally "Deep Squeak") is able to overcome this technical barrier. It was developed by researchers at the University of Washington School of Medicine.
According to experts, the program captures sound signals and transforms them into sonograms - computer images created using information about sound waves. This "translation" of acoustic data into visual data allows the use of modern machine vision algorithms developed for automatic control vehicles for analysis.
In other words, the new program for the first time allowed scientists to use artificial neural networks with deep learning to detect and analyze the "speech" of rodents.
As co-author Russell Marx explained, DeepSqueak uses biomimetic algorithms that learn to separate "unnecessary" sounds by relying on existing examples of animal vocalizations and extraneous noises.
Note that Marks and his colleague Kevin Coffey (Kevin Coffey) are studying the psychological effects of taking various drugs. In particular, researchers are interested in how the behavior of animals changes during stress and the formation of different types of addiction.
According to experts, the DeepSqueak program will help them find out what the rodents "say" to each other during experiments.
Promotional video:
“Animals have a rich set of [acoustic] signals, about 20 species. By misusing drugs, they emit both positive and negative signals,”said Kevin Coffey, talking about the complex nature of each addiction.
The scientist also shared an interesting discovery that the DeepSqueak program made possible. It turned out that male mice emit the same characteristic vocalizations when communicating with each other. However, when the female approaches, their sound signals change: they become more complex and resemble a "song" associated with courtship.
By the way, this effect is more dramatic when the male smells the potential bride, but does not see her. These data suggest that the repertoire of male mice contains specific "songs" for different stages of courtship, the researchers said.
In addition, it was found that rodents are more likely to make "happy" sounds when they play with relatives or expect a reward, for example, a piece of sugar.
However, behavioral and evolutionary biologists are more likely to be interested in such observations, and the team led by Professor John Neumaier is more concerned with the features associated with alcohol and opioid addiction (the latter is formed when taking various painkillers). According to experts, the improved analysis of the "speech" of rodents will help develop new methods of dealing with such ailments.
"If scientists can better understand how drugs alter brain activity to produce pleasure or discomfort, more effective treatments for addiction can be developed," the professor concluded.
A scientific article with a more detailed description of the new program was published in the journal Neuropsychopharmacology.
