Researchers from the University of Lund in Sweden have developed and tested a
new solution for processing and storing large amounts of data that are expected from future implantable neurocomputer interfaces. The system will simultaneously receive data from more than 1 million neurons in real time. After converting the data, they will be sent for processing and storage on ordinary computers. The system will provide feedback at speeds up to 25 milliseconds, stimulating up to 100 thousand neurons.
A new technology can be used to monitor the brain of paralyzed patients, including to track signs of epilepsy, and for real-time feedback to monitor robotic hands of paralyzed patients.
Today, technical and neurobiological advances in brain-computer interfaces impose increasing requirements on databases and software for their processing, especially when it comes to working with real-time data obtained from a large number of neurons. In order to cope with this problem, scientists have created a scalable software architecture for parallel recording and data processing using standard computers. The architecture has shown the ability to handle real-time information and provide reverse response at speeds of less than 25 milliseconds. Researchers are confident that their development will be suitable for working with both existing and future neurocomputer interfaces.
“A noticeable advantage of the architecture and data format is that the information does not require further translation, since the brain signals are translated directly into the code,” the researchers note. Thanks to this approach, an ordinary computer can work with data, and the processing speed is so high.
The real-time monitoring tool for large areas of the brain can be used for research, diagnosis and treatment. It should be especially effective for future implantable neurocomputer interfaces with feedback, which will help to monitor large areas in the brain of paralyzed patients, track incipient signs of epilepsy, and also to control paralyzed robotic arms.
The system is designed to register neural signals from implanted electrodes. Below is an example of such a device, an elastic biocompatible electrode
developed by scientists from Linköping University (Sweden), having a grid of 32 open metal contacts that, after implantation, come into contact with the brain tissue.
Below is a diagram of the system. The main clock (a) is synchronized with the receiving devices (b), which organize the
sorting of the spikes - register and classify the electrical activity of the neurons received from the subject (shown as a mouse face) (e), as well as data compression. Information is encoded in a data grid of time intervals. In HDF5 format, this mesh is sent to storage (d and f). The last two points are still being planned.
The existing solutions for recording brain activity are limited from 512 to 1024 channels, which makes it difficult to process and store them on personal computers. The maximum number of channels per subject was 1792, the indicator will continue to grow. DARPA is working in this direction: in 2016, the agency
launched a program to develop implantable neural interfaces to obtain "unprecedented signal resolution and bandwidth for transmitting information between the human brain and electronic systems." The interface should act as a "translator" between the neuron electrochemistry and the code available for processing by computers. As part of the Neural Engineering System Design program (NESD, “Designing Neuro-Engineering Systems”), the agency expects tools to be upgraded, including to compensate for the sight and hearing of patients: for example, visual information will have to be transmitted in digital form to the brain.
The next step was the conclusion by DARPA in 2017 of
contracts for the creation of high-resolution brain implants with five research organizations and one commercial company. Each of the channels combines information from tens of thousands of neurons, which gives a blurry and noisy picture with low resolution. The NESD program is designed to overcome this barrier.