Evaluating the DeepSqueak and Mouse Song Analyzer vocalization analysis systems in C57BL/6J, FVB.129, and FVB neonates

Abstract Background Communication is an essential behavior in mammals. Alterations in communication (neonatal crying) characterize numerous human neurodevelopmental conditions. Mice produce communicative vocalizations, known as ultrasonic vocalizations, (USVs) that can be recorded. The Mouse Song Analyzer is an automated USV analysis system while DeepSqueak is a semi-automated USV detection system. Method We used data from, C57BL/6J, FVB.129, and FVB neonates to compare the reliability of DeepSqueak and the Mouse Song Analyzer across various acoustic variables. Results We found that both systems detected a similar quantity of USVs for FVB.129 and FVB mice. However, DeepSqueak detected more USVs for C57BL/6J mice. High correlations were found between systems for each strain. When assessing duration, Deepsqueak detected USVs of a longer duration then the Mouse Song Analyzer across all strains. A low correlation between systems for duration was found for FVB.129 mice, while high correlations were found for C57BL/6J and FVB mice. When assessing fundamental frequency, the Mouse Song Analyzer detected a higher frequency than DeepSqueak for FVB.129 mice, with no other differences present. High correlations between systems were found for C57BL/6J and FVB.129 mice, while a low correlation was found for FVB mice. We also assessed each system’s sensitivity and found that Deepsqueak was able to detect softer USVs than the Mouse Song Analyzer. Conclusions These findings demonstrate that the strain of mouse used significantly affects the reliability of USV analysis systems. However, our data also indicates that DeepSqueak is more reliable and accurate than the Mouse Song Analyzer due to its increased sensitivity.