A system for automatic epilepsy diagnosis developed at Lobachevsky University
The machine learning model developed by Lobachevsky University researchers detects epilepsy based on brain rhythms with an accuracy of over 90%. This technology will help doctors assess the risk of patients developing seizures before they occur. In addition, this method can be used to analyse encephalograms of people in professions associated with high risk and responsibility, such as pilots, drivers, and law enforcement officers.
"An experienced doctor can identify the presence of pathology in a patient by examining the rhythms of electroencephalography (EEG). However, there are no strict criteria for diagnosing epilepsy. We attempted to address this issue and, using open databases, trained a model to differentiate between an 'epileptic' brain and a healthy one. This development shows the potential of automated EEG analysis for more rapid and precise epilepsy diagnosis, ultimately enhancing patient treatment outcomes," said Anton Malkov, the study's author and senior researcher at the Laboratory of Artificial Intelligence in Cardiology and Neuroscience at the UNN Research Centre for Artificial Intelligence.
Today, it takes hours of brain activity recording to diagnose epilepsy in patients using an electroencephalogram. However, the diagnosis is only made when a seizure is visible on the EEG, but in 30% of cases it is not possible to record convulsive activity electrographically, and the periods between seizures can be very long. To address this issue, scientists have proposed using interconvulsive brain activity to assess epilepsy.
"Standard clinical records often fail to take into account much of the valuable information contained in the EEG. Using the waves of brain rhythms, one can assess its condition, detect hyperexcitability, and identify abnormal brain function. Our model utilizes approximately 200 brain rhythm parameters to assess healthy or pathological activity," Anton Malkov explained.
"At this stage, the possibility of using background EEG activity to search for changes in brain function during the asymptomatic period has been demonstrated. This could help with early diagnosis, which is crucial for the treatment of conditions associated with brain disorders," noted Albina Lebedeva, co-author of the study and senior researcher at the Laboratory of Artificial Intelligence in Cardiology and Neuroscience at the UNN Research Centre for Artificial Intelligence.
It is planned to assess the model’s effectiveness using new clinical data at several Russian medical research centres. Furthermore, scientists intend to teach the system in the future to identify not just the presence but also the type of epilepsy based on EEG, which can point to its underlying causes.
The project was completed by the Laboratory of Artificial Intelligence in Cardiology and Neuroscience at the UNN Research Centre for Artificial Intelligence with the participation of the Institute of Theoretical and Experimental Biophysics of the Russian Academy of Sciences (Pushchino). The results have been published in the international journal Communications in Nonlinear Science and Numerical Simulation.
