Filtering Poor-Quality Heart Signals Improves Seizure Alerts

What was studied

Researchers looked at whether checking the quality of pulse signals from a wearable light sensor, called photoplethysmography (PPG), could improve seizure detection. PPG can estimate heart rate, but movement can make the signal noisy and less reliable.

They used data from the PROMISE home study in 28 children with refractory epilepsy who were being monitored for nighttime motor seizures with the NightWatch device. The analysis included 741 device alarms: 135 were true seizures and 606 were non-seizure events. The researchers compared three approaches: a new pulse-based quality assessment that removed poor-quality pulse signals before calculating heart rate, a simpler heart-rate-based quality check, and no quality check.

What they found

Both quality-check methods significantly improved seizure detection performance compared with using the heart-rate data without any quality check. The new pulse-based method showed greater improvement than the heart-rate-based method.

Across the tested detection settings, the pulse-based method increased sensitivity and also improved positive predictive value and F1-score on average. At sensitivity levels highlighted as clinically relevant in the abstract, the pulse-based method did better than the simpler heart-rate-based method, with relative improvements in F1-score of 4% to 15% and in positive predictive value of 5% to 19%.

In simple terms, removing low-quality PPG pulses before estimating heart rate was associated with more accurate seizure detection in this dataset.

Limits of the evidence

This was an analysis of data from one existing home-based study, not a new trial testing the upgraded method in real-time use. The study included 28 children with refractory epilepsy and focused on nocturnal motor seizures, so the results may not apply to adults, other seizure types, or daytime use.

The results are based on device alarms and statistical comparisons across chosen heart-rate thresholds. The abstract does not report how the method would affect battery use, comfort, or performance in other wearable devices. It also does not show whether this software change would improve outcomes for every patient at home.

For families and caregivers

This study suggests that filtering out poor-quality pulse data may help heart-rate-based seizure-detection wearables perform better, without adding new hardware.

For families, that may matter because wearable devices are often affected by movement and false alarms. Better signal checking could make heart-rate-based seizure alerts more dependable, but this study does not show that any one device will be accurate enough for all children or all seizure types.

What to watch next

Useful next steps would include testing this pulse-quality method prospectively in larger real-world groups and in different ages, seizure types, and wearable devices.

Terms in this summary

photoplethysmography (PPG)

A light-based sensor used in many wearables to estimate pulse and heart rate.

heart rate (HR)

How many times the heart beats in one minute.

sensitivity

The proportion of true seizures that the device correctly detects.

positive predictive value (PPV)

The proportion of device alarms that are actually true seizures.

F1-score

A single measure that combines how well a test finds true events and avoids false alarms.

refractory epilepsy

Epilepsy that does not respond well to treatment.