Researchers from St Petersburg have developed an artificial intelligence model capable of analyzing the activity of brain neurons. In the near future, the technology could help evaluate the effectiveness of new drugs for neurodegenerative diseases and shed light on the causes of Alzheimer’s disease, they said. The findings were published in the journal IEEE Transactions on Neural Systems and Rehabilitation Engineering.
“We have developed a universal AI model called NEuRT for analyzing the activity of neural networks in the brain. Figuratively speaking, the neural network allows us to observe how neurons ‘talk’ to each other,” said Yekaterina Pchitskaya, head of the Biomedical Image and Data Analysis Laboratory at Peter the Great St Petersburg Polytechnic University (SPbPU).
The neural network is based on an architecture originally designed for analyzing human language, the researchers explained.
“It turned out that neuron signals and text are organized in a similar way – in both cases, sequence and context are crucial. The model was trained on the principle of ‘guess the missing word’, but instead of words, it reconstructed hidden fragments of neural recordings,” Pchitskaya said.
The neural network was trained on a vast open dataset of 270 gigabytes of neuronal recordings from the mouse visual cortex, the scientists said. The model then successfully applied what it had learned to another brain region – the hippocampus. The researchers recorded neural activity in mice using miniature fluorescence microscopy: tiny devices attached to the animal’s head allow observation of neurons while the mouse moves freely.