Sonified EEG could be useful triage tool

 

Medical students and nurses who listened to 15 seconds of single-channel sonified electroencephalograms detected seizures with 95%-98% sensitivity, outperforming neurologists who reviewed traditional visual EEG displays, according to the results of a single-center study.

“We confirm that individuals without EEG training can detect ongoing seizures or seizurelike rhythmic and periodic patterns by merely listening to short clips of sonified EEG,” wrote Josef Parvizi, MD, PhD, and his associates at Stanford (Calif.) University. “Ours is also the first study to test the capability of a sonification method to detect a range of significant abnormalities when it is used by clinical staff (e.g., physicians, nurses, and students).” The findings were published online March 20 in Epilepsia.

chaikom/iStock/Getty Images

The sonification technique is based on a new algorithm that translates low-frequency EEG signals into “speech-like declamations,” the investigators said. Vocal pitch, loudness, and resonance vary depending on input. Unlike prior sonification methods, brain rhythms, rate, and seizure severity are conserved.

To test the method, 34 medical students and 30 nurses watched a 4-minute training video before listening to 84 sonified EEGs: seven seizures, 52 slowing or normal patterns, and 25 seizurelike abnormalities (generalized periodic discharges, lateralized periodic discharges, triphasic waves, or burst suppression). For each patient, listeners heard two sonified EEG clips, one from each hemisphere, and designated them as “seizure,” “nonseizure,” or “don’t know.” For comparison, 12 EEG-trained neurologists and 29 EEG-trained medical students reviewed traditional visual displays of the same EEGs.

Sonified EEGs identified seizures with a sensitivity of 95% (standard deviation, 14%) when heard by nurses and 98% (SD, 5%) when heard by medical students. In contrast, the sensitivity of visual displays was only 88% (SD, 11%) when reviewed by neurologists and 76% (SD, 19%) when reviewed by EEG-trained medical students. Specificity of sonified EEGs was 85% when heard by the medical students and 82% when heard by the nurses. Specificity of traditional review was 90% for neurologists and 65% for medical students.

The study was based on a representative sample, not a prospectively and consecutively recruited cohort, which constrains insights into how this technique might perform “at the bedside,” the researchers said. Additionally, the sonification method would not identify focal seizures occurring outside the individual channels selected (in this study, T3-T5 or T4-T6).

The study was funded by a Stanford University BioX Seed Grant. Dr. Parvizi and one coinvestigator invented the sonification method described and cofounded a startup that has licensed the technology from Stanford University. The other two investigators had no conflicts.

SOURCE: Parvizi J et al. Epilepsia. 2018 Mar 20. doi: 10.1111/epi.14043.

 

Medical students and nurses who listened to 15 seconds of single-channel sonified electroencephalograms detected seizures with 95%-98% sensitivity, outperforming neurologists who reviewed traditional visual EEG displays, according to the results of a single-center study.

“We confirm that individuals without EEG training can detect ongoing seizures or seizurelike rhythmic and periodic patterns by merely listening to short clips of sonified EEG,” wrote Josef Parvizi, MD, PhD, and his associates at Stanford (Calif.) University. “Ours is also the first study to test the capability of a sonification method to detect a range of significant abnormalities when it is used by clinical staff (e.g., physicians, nurses, and students).” The findings were published online March 20 in Epilepsia.

chaikom/iStock/Getty Images

The sonification technique is based on a new algorithm that translates low-frequency EEG signals into “speech-like declamations,” the investigators said. Vocal pitch, loudness, and resonance vary depending on input. Unlike prior sonification methods, brain rhythms, rate, and seizure severity are conserved.

To test the method, 34 medical students and 30 nurses watched a 4-minute training video before listening to 84 sonified EEGs: seven seizures, 52 slowing or normal patterns, and 25 seizurelike abnormalities (generalized periodic discharges, lateralized periodic discharges, triphasic waves, or burst suppression). For each patient, listeners heard two sonified EEG clips, one from each hemisphere, and designated them as “seizure,” “nonseizure,” or “don’t know.” For comparison, 12 EEG-trained neurologists and 29 EEG-trained medical students reviewed traditional visual displays of the same EEGs.

Sonified EEGs identified seizures with a sensitivity of 95% (standard deviation, 14%) when heard by nurses and 98% (SD, 5%) when heard by medical students. In contrast, the sensitivity of visual displays was only 88% (SD, 11%) when reviewed by neurologists and 76% (SD, 19%) when reviewed by EEG-trained medical students. Specificity of sonified EEGs was 85% when heard by the medical students and 82% when heard by the nurses. Specificity of traditional review was 90% for neurologists and 65% for medical students.

The study was based on a representative sample, not a prospectively and consecutively recruited cohort, which constrains insights into how this technique might perform “at the bedside,” the researchers said. Additionally, the sonification method would not identify focal seizures occurring outside the individual channels selected (in this study, T3-T5 or T4-T6).

The study was funded by a Stanford University BioX Seed Grant. Dr. Parvizi and one coinvestigator invented the sonification method described and cofounded a startup that has licensed the technology from Stanford University. The other two investigators had no conflicts.

SOURCE: Parvizi J et al. Epilepsia. 2018 Mar 20. doi: 10.1111/epi.14043.

 

Medical students and nurses who listened to 15 seconds of single-channel sonified electroencephalograms detected seizures with 95%-98% sensitivity, outperforming neurologists who reviewed traditional visual EEG displays, according to the results of a single-center study.

“We confirm that individuals without EEG training can detect ongoing seizures or seizurelike rhythmic and periodic patterns by merely listening to short clips of sonified EEG,” wrote Josef Parvizi, MD, PhD, and his associates at Stanford (Calif.) University. “Ours is also the first study to test the capability of a sonification method to detect a range of significant abnormalities when it is used by clinical staff (e.g., physicians, nurses, and students).” The findings were published online March 20 in Epilepsia.

chaikom/iStock/Getty Images

The sonification technique is based on a new algorithm that translates low-frequency EEG signals into “speech-like declamations,” the investigators said. Vocal pitch, loudness, and resonance vary depending on input. Unlike prior sonification methods, brain rhythms, rate, and seizure severity are conserved.

To test the method, 34 medical students and 30 nurses watched a 4-minute training video before listening to 84 sonified EEGs: seven seizures, 52 slowing or normal patterns, and 25 seizurelike abnormalities (generalized periodic discharges, lateralized periodic discharges, triphasic waves, or burst suppression). For each patient, listeners heard two sonified EEG clips, one from each hemisphere, and designated them as “seizure,” “nonseizure,” or “don’t know.” For comparison, 12 EEG-trained neurologists and 29 EEG-trained medical students reviewed traditional visual displays of the same EEGs.

Sonified EEGs identified seizures with a sensitivity of 95% (standard deviation, 14%) when heard by nurses and 98% (SD, 5%) when heard by medical students. In contrast, the sensitivity of visual displays was only 88% (SD, 11%) when reviewed by neurologists and 76% (SD, 19%) when reviewed by EEG-trained medical students. Specificity of sonified EEGs was 85% when heard by the medical students and 82% when heard by the nurses. Specificity of traditional review was 90% for neurologists and 65% for medical students.

The study was based on a representative sample, not a prospectively and consecutively recruited cohort, which constrains insights into how this technique might perform “at the bedside,” the researchers said. Additionally, the sonification method would not identify focal seizures occurring outside the individual channels selected (in this study, T3-T5 or T4-T6).

The study was funded by a Stanford University BioX Seed Grant. Dr. Parvizi and one coinvestigator invented the sonification method described and cofounded a startup that has licensed the technology from Stanford University. The other two investigators had no conflicts.

SOURCE: Parvizi J et al. Epilepsia. 2018 Mar 20. doi: 10.1111/epi.14043.