A neural network for error-free detection of epilepsy developed in Russia

Non-convulsive epileptic seizures canhappen unexpectedly, so scientists are looking for ways to best predict them. Researchers from St Petersburg University have made great progress in this - they have developed a neural network to detect such seizures based on electroencephalography (EEG) data.

Seizures happen because of a sudden electrical discharge in the neurons of the brain, which is why they are so difficult to predict in advance.

To test the neural network developed by specialists, they loaded into it the data of 83 patients from the N.M. Pirogov. After that, her findings were compared with the diagnoses of real doctors.

It is reported that the results of the algorithm and physicians coincided in 80% of cases. When the scientists ruled out 23 low-quality EEG recordings, they found that the computer's accuracy reached 100%.

Earlier, Innopolis University also developed an algorithm with which scientists trained a neural network to predict epilepsy attacks.

Source: gazeta