A robust DeepLabCut-based pipeline for the multi-dimensional quantification of vestibular dysfunction in mice.

Abstract

Accurate quantification of behavioral deficits is critical for investigating vestibular disorders, yet current assessment methods often fail to capture complex kinematic features like rotational asymmetry. This study aimed to develop a cost-effective, high-precision automated analysis pipeline using the pose estimation tool DeepLabCut to quantify fine-grained behavioral phenotypes in a mouse model of Bilateral Vestibular Dysfunction (BVD). We established the BVD model via bilateral intratympanic gentamicin injection and monitored locomotor function using the Open Field Test over a 7-day post-operative period. By tracking anatomical landmarks (nose, head, and tail) with DeepLabCut and utilizing a custom Python analysis framework, we quantified total locomotor distance, central zone exploration, and specific rotational metrics. The analysis revealed a distinct biphasic behavioral progression: an acute phase (Days 1-2) characterized by significant hypokinesia and thigmotaxis, indicative of severe spatial disorientation, followed by a chronic phase (Days 5-7) marked by a transition to profound hyperactivity and pathological, stereotypic circling. Crucially, the pipeline detected early onset rotational tendencies that traditional metrics overlooked. This study demonstrates that the proposed DeepLabCut-based methodology provides a sensitive, objective, and accessible tool for the multi-dimensional assessment of vestibular dysfunction, offering a robust paradigm for evaluating functional recovery and therapeutic interventions.