House mice and many other species use ultrasonic vocalizations (USVs) to communicate in various contexts including social and sexual interactions. These vocalizations are increasingly investigated in research on animal communication and as a phenotype for studying the genetic basis of autism and speech disorders. Because manual methods for analyzing vocalizations are extremely time consuming, several methods have been recently developed for automatically detecting and classifying USVs. Abbasi and colleagues evaluated the advantages and disadvantages of these methods in a full, systematic comparison, while also presenting a new approach.
They found two tools outperformed others in terms of correct detections, and one of these, Automatic Mouse Ultrasound Detector (or A-MUD), also outperformed the others for minimizing false detections. A-MUD performed as well as manual detection and it does not require any parameter tuning or custom training of the networks. Their new method for automating USV classification, called BootSnap, combines bootstrapping on Gammatone Spectrograms and Convolutional Neural Networks algorithms with Snapshot ensemble machine learning (see schematic diagram). BootSnap outperformed the pretrained and retrained state-of-the-art tool, and thus it is more generalizable. BootSnap is free for scientific use.
To evaluate the current methods and develop new tools, Abbasi collaborated with an interdisciplinary team of researchers of the Konrad Lorenz Institute of Ethology (University of Veterinary Medicine, Vienna, Austria), who study the vocalizations of house mice. In the paper, they pointed out that the generalizability of methods for detection and classification of mouse vocalizations, which is often overlooked, is of great importance and provided a solution for that.
The results are published in PLOS Computational Biology. This open access scientific journal features works of exceptional significance that further our understanding of living systems at all scales—from molecules and cells, to patient populations and ecosystems—through the application of computational methods.
Abbasi, R., Balazs, P., Marconi, M.A., Nicolakis, D., Zala, S.M., Penn, D.J. (2022): "Capturing the songs of mice with an improved detection and classification method for ultrasonic vocalizations (BootSnap)" in: PLOS Computational Biology
doi.org/10.1371/journal.pcbi.1010049
