Comprehensive Summary
This study looked at a new computer system called IEDD that helps detect abnormal brain signals, called interictal epileptiform discharges (IEDs), which are important in diagnosing epilepsy. Normally, doctors must review EEGs by hand, which is slow and not always consistent. The researchers combined two kinds of AI (convolutional networks and Transformers) and added a data-generating tool (GAN) to boost the training data. They tested the system on EEGs from 11 children (1,206 IED events) and 13 infants with spasms. The model reached 96.1% accuracy in the first group and 95.3% accuracy in the second. In more detailed testing of different IED types, it showed an average sensitivity of 87.3% and precision of 90%, with the best results for detecting spikes (95.9% sensitivity).
Outcomes and Implications
Catching IEDs quickly and correctly matters because missed or false alarms can change an epilepsy diagnosis or treatment plan. This system showed it can spot these events with both high accuracy and high confidence, which could save doctors hours of manual review while reducing mistakes. For patients, that means a better chance of getting the right diagnosis and treatment sooner. While it is not yet ready to use at the bedside, with larger studies and real-world testing, it could become part of routine EEG monitoring in epilepsy clinics within the next few years.