Sounds of seizures.

Area of Science:

• Neurology
• Biomedical Engineering
• Medical Acoustics

Background:

• Epilepsy monitoring units (EMUs) are crucial for seizure diagnosis.
• Objective seizure detection methods are needed to supplement clinical observation.
• Audio analysis presents a novel approach for seizure detection.

Purpose of the Study:

• To determine the accuracy of identifying seizures using audio recordings.
• To assess the feasibility of audio-based seizure detection.
• To guide the development of audio seizure detection technology.

Main Methods:

• Generated 166 audio clips (30s) from 83 epilepsy patients (seizure vs. non-seizure).
• Five epileptologists performed blinded reviews of audio clips.
• Calculated accuracy, positive predictive value (PPV), and negative predictive value (NPV) for consensus ratings.

Main Results:

• Consensus ratings achieved PPV of 0.91 and NPV of 0.66.
• High confidence ratings improved performance (PPV 0.96, NPV 0.70).
• Hyperkinetic and tonic-clonic seizures were accurately identified; non-motor seizures were not.

Conclusions:

• Epileptologists can accurately identify sound-producing seizures from audio.
• Specific sounds like disordered breathing and vocalizations aid identification.
• Findings support the development of audio-based seizure detection devices.