Event-Related Potential Interpretation Approaches for Neonatal Hearing Screening

Abstract

Biopotentials are the electrical discharges generated from the human body. They reflect the status underlying the phenomenon and are frequently used to assess the real-time physiological status of the human body. Electroencephalography (EEG), Electrocorticography (ECoG), Electrocardiography (ECG), Electromyography (EMG), and Electrooculography (EOG) assess the health status of the brain, heart, muscles, eyes, and nerve cells. This research utilizes stimuli-evoked EEGs to draw insights into the auditory pathway of newborns. The magnitude of the neural component of the acquired human EEG is in the range of a few microvolts; hence, it often gets affected by physiological and non-physiological artifacts. Therefore, it is crucial to understand the difference between neural and non-neural components of acquired EEGs. A pilot study of epilepsy seizure detection and classification was performed by interpreting the actual human subjects' multichannel EEG (n=88). Developed EEG interpretation algorithm could delineate between normal and epileptic subjects with no false positives. Additionally, for epileptic patients, the seizure detection classifier resulted in 93.18 % accuracy, while results were compared with clinical impressions provided by experienced neurophysiologists. Moreover, the blind validation study confirmed the generalizability of the developed algorithm. The pilot study on epileptic seizure detection and classification helped obtain accurate neural inferences from human EEGs. The next step was to develop and validate a stimuli-evoked Event-Related Potential (ERP) extraction system. Before deploying the developed system on newborns for the final objective, we validated the developed system by conducting experiments on young adults. The competency of the developed ERP extractor was tested by two different experimental paradigms for five young adults: (i) auditory evoked brainstem response (ABR, brainstem response), and (ii) Mismatch Negativity (MMN; cortical response). The obtained responses were validated against CE-certified acquisition systems. Validation showed that the grand average response patterns for n=5 subjects matched. Moreover, ERP image plots analysis showed intertrial variabilities, which can be further used for test time reduction. The obtained results encouraged us to proceed with neonatal experiments. A comfortable headband was developed, and experienced neonatologists assessed the suitability of the developed headband before conducting experiments. ABR is the current gold standard for newborn hearing screening. The developed system replicates the current gold standard (ABR) and enhances the existing neonatal hearing screening by extracting an additional cortical auditory evoked potential (MMN), scanning the complete auditory pathway. Brainstem (ABR) and cortical (MMN) responses from three neonates were obtained using the developed system, and the response showed characteristic peaks within expected temporal latencies. MMN and ABR response extraction and interpretation are in progress as a final validation process. We envisage acquiring ABR and MMN data from more subjects to include more neonatal EEG variabilities to develop a neonatal hearing screening system providing auditory evoked brainstem (ABR) and cortical (MMN) response extraction.