Validating an Automated Rodent Ultrasonic Vocalization Data Collator
Abstract
Rodent ultrasonic vocalizations (USVs) are a key dependent measure within behavioral science that have translational relevance with respect to medical, especially neuropsychiatric, illnesses. In recent years, several programs have been developed to enable the recording, labeling, and parameterization of rodent USVs via conventional, machine learning, or deep learning approaches. However, there remains a need to further automate the downstream process of data collation so that labeled and classified USV parameters can be rapidly formatted and analyzed via statistical approaches based on experimenter preferences. The aim of this thesis was therefore to validate an automated vocalization data collator (AVDC) to expedite the collation of data from rodent USV studies, particularly using UltraVox XT. This was accomplished by comparing USV data outputs from an AVDC and manual vocalization data collation pipeline (MVDC) twice. Results revealed some discrepancies between the two approaches when examined in a qualitative and semi-quantitative manner; however, there were no statistically significant differences between the performance of the AVDC and MVDC, indicating that the AVDC is fit for purpose. While there are some limitations which researchers should be aware of with respect to the AVDC, this work is relevant to accelerating basic and pre-clinical investigations involving rat USVs.