When it comes to diagnosing neurological disorders, it is important for a physician not only to obtain a result, but also to understand why the system arrived at that specific conclusion. This is precisely the problem solved by specialists from the Plekhanov Russian University of Economics and the Pirogov National Medical and Surgical Center: they developed an EEG analysis algorithm that not only detects seizures with high accuracy, but also explains its decisions.
The main feature of the new method is the use of "explainable" artificial intelligence adapted for electroencephalography. Usually, neural networks work like a "black box": the physician sees the final result but does not understand what it is based on.
However, in this case, the algorithm shows exactly which areas of the cerebral cortex and which frequency ranges were decisive for the diagnosis. The system displays this in the form of heat maps.
For medicine, this is important for several reasons at once. First, the physician can verify the AI's conclusions and trust the result more. Second, the method confirmed known neurophysiological mechanisms and identified new biomarkers. For example, this concerns the high-frequency gamma range, which was previously considered interference.
The developers believe that in the future this approach can be adapted for diagnosing other neurological and neurodegenerative diseases.
Read more on the topic:
- Without surgeries and implants: a new way to control brain neurons found in Russia
- Russian Scientists Learn to "See" Blood Flow in Deep Brain Arteries in Real Time
- AI for brain tumor diagnostics created by scientists from UNN