Validating the Gap-Startle Paradigm for Tinnitus Detection: A Machine Learning Approach in CBA/CaJ Mice

Tinnitus, a prevalent hearing disorder affecting one-third of Americans, manifests as a buzzing or ringing sensation in the ears without an external acoustic stimulus. Gap prepulse inhibition of the acoustic startle reflex (GPIAS) serves as an objective method for tinnitus screening in rodents, wherein a diminished reflex signifies the primary pitch of tinnitus, as tinnitus fills the gap.

However, Lobarinas et al. (2013) highlighted a critical limitation: rodents often exhibit reduced acoustic startle reflexes after acoustic trauma or hearing loss, creating a floor effect that renders further suppression undetectable even if gap perception persists, resulting in false-positive tinnitus screening outcomes.

To overcome this hurdle, we employed the CBA/CaJ mouse model to evaluate the efficacy of tactile airpuff modification and utilized a machine learning algorithm to classify startle responses, achieving 98% accuracy in distinguishing startles from non-startles. By inducing unilateral conductive hearing loss through ear plugging, we observed enhanced gap detection ability, contrasting with the false-positive tinnitus indicators reported by Lobarinas et al.

We also induced tinnitus pharmacologically using sodium salicylate, uncovering frequency-specific alterations in gap detection patterns. Our findings suggest that variations in data analysis methodologies, particularly the use of machine learning algorithms to filter out non-startle responses, may account for species-specific disparities between mouse and rat models.

Moreover, these insights could significantly bolster the validity of GPIAS as a tinnitus screening assessment tool.