Machine Learning Tool Shows Promise for Detecting EEG Spikes

Source: Journal of neurology

Summary

This study looked at how well a machine learning software called P15 can detect interictal epileptiform discharges (IEDs) in EEG recordings from patients at a large hospital in England. IEDs are important for diagnosing and classifying epilepsy. The researchers analyzed EEGs collected from routine clinical practice between June 2024 and February 2025, comparing the software's results to the final clinical reports made by human experts.

The key findings showed that the P15 software was able to correctly identify IEDs in about 81% of cases where they were present, which is a good level of sensitivity. However, it also had a lower specificity of about 60%, meaning it sometimes incorrectly identified IEDs when they weren't there. The software was very good at ruling out IEDs, with a negative predictive value of 96%, but it had a low positive predictive value of about 20%, indicating that it may mistakenly flag many EEGs as having IEDs.

These results are important because they suggest that the P15 software could be a helpful tool for doctors to rule out IEDs in patients, potentially making the diagnosis process faster. However, the low positive predictive value raises concerns about its reliability in identifying IEDs accurately. More research is needed to improve the software and understand how it can be used effectively alongside clinical judgment before it can be widely adopted in patient care.

Original source