A novel system using deep learning technology and custom electroencephalogram (EEG) data techniques has shown that it automatically learns seizure signatures in patients with epilepsy and reduces the amount of raw EEG data reviewed by a neurologist must, according to a study published in EBioMedicine.
The researchers involved in the study obtained scalp EEG data from 365 patients with 171,745 sictal and 2,185,864 sectal samples. The data was analyzed using artificial intelligence (AI) during a crowdsourcing challenge.
The study participants were asked to develop deep learning models for the automatic annotation of epileptic seizures in the raw EEG data. Specifically, these participants worked on the development of an ictal / interictal classifier that showed high sensitivity and low false alarm rates. A challenge platform was then set up to prevent participants from downloading or accessing the relevant data.
Overall, the results showed that the automatic detection system had adjustable sensitivities from 75.00% to 91.60%. This achievement with high sensitivity rates reduced the amount of raw EEG data that had to be checked by a human annotator by the maximum achievable reduction factor of 142x or 22x.
The study’s researchers stated that the algorithm allows for “instant optimization of the balance between sensitivity” and the amount of raw EEG data that needs to be verified.
Ultimately, the deep learning technique enabled participants to learn patient-specific seizure signatures. The study researchers added that the system can then “filter out seizure segments from raw EEG data for review by an experienced neurologist.”
The primary limitation of this study was the relatively small sample size. Further research is needed to validate the results of this study.
In addition to seizure detection, the study’s investigators concluded that “the platform enables collaboration between data scientists while keeping proprietary or sensitive data safe and secure”.
Disclosure: This investigation was partially supported by IBM. For a full list of specifications, see the original reference.
S. Roy, I. Kiral, M. Mirmomeni et al .; IBM Epilepsy Consortium. Evaluation of artificial intelligence systems to assist neurologists with quick and accurate annotations of electroencephalography data of the scalp. EBioMedicine. 2021; 66: 103275. doi: 10.1016 / j.ebiom.2021.103275