CortixAI: A Low-Cost EEG-Based Automated Hearing Screening System Using Consumer-Grade Hardware and Machine Learning

Abstract

This paper presents CortixAI, a low-cost automated hearing screening system that achieves clinical-grade accuracy using consumer–grade electroencephalography (EEG) hardware and machine learning. The system integrates an OpenBCI Cyton biosensing platform with a custom spring-loaded Ag/AgCl electrode array mounted on a lightweight, flexible cable sleeve framework and bone-conduction stimulus delivery. This architecture enables nonspecialist operation without scalp preparation or abrasive electrode gel. A hybrid classifier combining a 1D convolutional neural network and linear discriminant analysis achieved 91.7% balanced accuracy in the top-responder subgroup and 87.4% sensitivity with 96.0% specificity across a 30-participant pilot cohort under leave-one-participant-out cross-validation. The complete system bill of materials totals under $500 USD, representing a roughly 40-fold cost reduction relative to commercial Automated Cortical Auditory Evoked Potential (ACAEP) systems. CortixAI addresses a critical gap in global hearing healthcare, where the high cost and operator-skill requirements of clinical equipment have prevented routine objective screening from reaching the populations that would benefit most from early intervention.