MOSCOW, June 13 A program for diagnosing epilepsy, which helps to reduce the routine work of a doctor by 95%, was developed by IKBFU specialists. I.Kant together with other Russian scientists. According to them, the development has no direct analogues in the world. The results are published in Applied Science.
Epilepsy is a dangerous neurological disease that, according to WHO, affects up to 50 million people worldwide. Its main symptoms are the so-called seizures, in which there is a pathological synchronization of brain neurons. As experts explained, often seizures go away without external signs, so many people may not suspect that they are sick.
To diagnose and treat this disease in clinical practice, electroencephalography (EEG) is used, which allows recording the electrical activity of the brain. Existing methods involve «manual» processing of EEG results by a doctor, which can take up to six hours for one patient. Kant Institute (I. Kant Baltic Federal University), together with specialists from the National Medical and Surgical Center named after A. N.I. Pirogov developed a program that can independently search in the EEG data array for signs of abnormal electrical activity characteristic of a seizure state. According to the creators, the proposed algorithm will become the basis of the medical decision support system (DMSS).
«
"SPPVR conducts preliminary marking of EEG signals and highlights fragments with suspicion of epileptic seizures, which are then evaluated by an epileptologist. The system can significantly reduce the doctor’s workload by reducing the EEG processing time to 10–20 minutes,” said co-author of the study, chief researcher at the Baltic Center for Neurotechnologies and Artificial Intelligence of IKBFU. I. Kant Alexander Khramov.
The new program, as the scientists explained, is based on machine learning without a «teacher» — unlike neural network systems, it independently reveals patterns and hidden relationships in unlabeled data without relying on pre-set templates. This approach to diagnosing epilepsy, according to the authors, was used by them for the first time in the world.
«We do not «train» the algorithm in advance on some data structure, but give it freedom in the analysis. Using this approach we automatically highlight special signals that differ from normal brain activity, which most often correspond to epileptic seizures,» Khramov explained.
< /span>
This principle, according to scientists, makes the process of making a diagnosis more transparent and controllable. In addition, it reduces the risk of diagnostic errors caused by both the human factor and the incorrect operation of the program, since the final decision remains with the epileptologist.
In the future, the scientific team intends to adapt the proposed program for combined use with the method of photostimulation of the patient's brain allowing to significantly reduce the time required for a full-fledged examination using the EEG. /20230504/chatboty-1869425425.html» data-title='»Just ask me». Scientists evaluated the recommendations of medical chatbots'>
The research is being carried out as part of a strategic project of the IKBFU. I. Kant «Cognitive longevity» of the state program for supporting universities «Priority-2030».