RAINSTORM: Automated Analysis of Mouse Exploratory Behavior Using Artificial Neural Networks

Abstract

Rodent exploratory behavior is widely used for assessing cognitive function. RAINSTORM—Real and Artificial Intelligence for Neuroscience-Simple Tracker for Object Recognition Memory—is a versatile tool designed to streamline the analysis of such behaviors in rodents. This pipeline integrates manual behavioral scoring, geometric analysis, and artificial intelligence–powered behavioral labeling, offering reproducible, scalable, and efficient evaluation methods. RAINSTORM processes raw positional data and automates the identification of exploratory behaviors, providing insights into memory performance. This tool is designed to learn from the labeling criteria of one or more experimenters by capturing the different aspects of expert opinion and reducing subjective bias in subsequent scoring procedures. The experimenter can go from unprocessed pose estimation data (coordinates of the animal´s key body parts obtained with open-source software such as DeepLabCut) to accurate exploration patterns in a matter of minutes. By optimizing the analysis process, RAINSTORM significantly enhances the reliability and efficiency of behavioral research. RAINSTORM has also become a robust methodology for assessing recognition memory in rodents by accurately quantifying exploration times for familiar and novel objects, and it has been extended to include (but is not limited to) a wider range of exploratory behaviors. The software is readily applicable to other experimental designs that rely on quantifying exploration in mice, such as social preference and object pattern separation, among others.