Automated spike and seizure detection: Can we use it? Using automated spike and seizure detection when reviewing prolonged EEGs at the Epilepsy Monitoring Unit

Reus, E.E.M. (2024)

Doctoral Thesis

The entire healthcare system is grappling with shortages of personnel that are likely to grow. There are increasingly more people in need of care, but the number of people providing care cannot keep up.
This thesis examines whether the assessment of long-term EEG registrations (4 – 90 hours) in adults can be partially taken over by an algorithm, referred to as automated spike and seizure detection software in the thesis. This is done without compromising the quality of the assessment.
We evaluated the performance of three commercially available software packages (Persyst, Encevis, and BESA) to determine if they perform as well as human experts. We looked at both interictal and ictal abnormalities (epileptic seizures).
In conclusion, the software detects interictal abnormalities as well as human experts do. However, epileptic seizures are not detected as well as by human experts, especially for seizures with short duration (less than 10 seconds) or subtle EEG abnormalities. Overall, the Persyst software package performs the best.
Furthermore, this thesis includes a new method in which automatic detection is combined with sampled visual assessment. We compared this new assessment method with the conventional assessment method (based on complete visual assessment) and found no difference in conclusion.

• Automated spike detection
• Automated seizure detection
• EEG
• EMU
• Epilepsy

• All authors: Reus, E.E.M.
• Supervisor: Dijk, J.G van
• Co-supervisor: Visser, G.V.; Cox, F.M.E.
• Committee: Lammers, G.J.; Zijlmans, G.J.M.; Alfen, N. van; Thijs, R.D.
• Qualification: Doctor (dr.)
• Awarding Institution: Faculty of Medicine, Leiden University Medical Center (LUMC), Leiden University
• Date: 2024-01-30
• ISBN (print): 9789083374383