Abstract:Objective: The traditional approach to interpreting electroencephalograms (EEGs) requires physicians with formal training to visually assess the waveforms. This approach can be less practical in critical settings where a trained EEG specialist is not readily available to review the EEG and diagnose ongoing subclinical seizures, such as nonconvulsive status epilepticus. Methods: We have developed a novel method by which EEG data are converted to sound in real time by letting the underlying electrophysiological signal modulate a voice tone that is in the audible range. Here, we explored whether individuals without any prior EEG training could listen to 15-second sonified EEG and determine whether the EEG represents seizures or nonseizure conditions. We selected 84 EEG samples to represent seizures (n = 7), seizure-like activity (n = 25), or nonperiodic, nonrhythmic activity (normal or focal/generalized slowing, n = 52). EEGs from single channels in the left and right hemispheres were then converted to sound files. After a 4-minute training video, medical students (n = 34) and nurses (n = 30) were asked to designate each audio sample as "seizure" or "nonseizure." We then compared their performance with that of EEG-trained neurologists (n = 12) and medical students (n = 29) who also diagnosed the same EEGs on visual display. Results: Nonexperts listening to single-channel sonified EEGs detected seizures with remarkable sensitivity (students, 98% ± 5%; nurses, 95% ± 14%) compared to experts or nonexperts reviewing the same EEGs on visual display (neurologists, 88% ± 11%; students, 76% ± 19%). If the EEGs contained seizures or seizure-like activity, nonexperts listening to sonified EEGs rated them as seizures with high specificity (students, 85% ± 9%; nurses, 82% ± 12%) compared to experts or nonexperts viewing the EEGs visually (neurologists, 90% ± 7%; students, 65% ± 20%). Significance: Our study confirms that individuals without EEG training can detect ongoing seizures or seizure-like rhythmic periodic patterns by listening to sonified EEG. Although sonification of EEG cannot replace the traditional approaches to EEG interpretation, it provides a meaningful triage tool for fast assessment of patients with suspected subclinical seizures.

On sound-based interpretation of neonatal EEG

Neural Tracking of Sound Rhythms Correlates with Diagnosis, Severity, and Prognosis of Disorders of Consciousness

Detecting silent seizures by their sound

A method for AI assisted human interpretation of neonatal EEG

Neonatal EEG Interpretation and Decision Support Framework for Mobile Platforms

A Method to Study Global Spatial Patterns Related to Sensory Perception in Scalp EEG.

Rapidly learned identification of epileptic seizures from sonified EEG

Application of Convolutional Neural Network for Decoding of 12-Lead Electrocardiogram from a Frequency-Modulated Audio Stream (Sonified ECG)

Automatic detection and sonification of nonmotor generalized onset epileptic seizures: Preliminary results

Delving into α-stable distribution in noise suppression for seizure detection from scalp EEG.

An EEG analysis framework through AI and sonification on low power IoT edge devices

Neonatal Frequency-Following Responses: A Methodological Framework for Clinical Applications

Electrographic monitoring for seizure detection in the neonatal unit: current status and future direction

Amplitude-Integrated Electroencephalography Applications And Algorithms In Neonates: A Systematic Review

NeoGuard: a public, online learning platform for neonatal seizures

Cortical Auditory-Evoked Responses in Preterm Neonates: Revisited by Spectral and Temporal Analyses

Deep Learning for EEG Seizure Detection in Preterm Infants

A Comparison of Automatically Extracted Quantitative EEG Features for Seizure Risk Stratification in Neonatal Encephalopathy

System Level Framework for Assessing the Accuracy of Neonatal EEG Acquisition

A rhythmic encoding approach based on EEG time-frequency image for epileptic seizure detection