Scientists develop prosthesis to bypass brain damage by re-encoding memories
Researchers at the University of Southern California (USC) and Wake Forest Baptist Medical Center have developed a neuroprosthetic device designed to help people with memory loss. The implantable system, based on a small array of electrodes and an advanced computational algorithm, has shown promising results in laboratory studies and is now being evaluated in human patients.
The device was originally developed at USC and further tested at Wake Forest Baptist. It builds on decades of foundational work by Ted Berger of the USC Viterbi School of Engineering and on a new algorithm created by Dong Song, also of USC Viterbi. Critical neural data for constructing and validating the models were gathered through a long-term collaboration with Sam Deadwyler and Robert Hampson of Wake Forest Baptist’s Department of Physiology & Pharmacology.
How sensory signals become memories
When the brain receives sensory input, that information is initially encoded as a complex pattern of electrical activity in the hippocampus, a brain region essential for forming memories. As the signal travels through multiple subregions of the hippocampus, it is re-encoded at each stage so that, by the time it reaches the final hippocampal region, it has been transformed into a pattern suitable for long-term storage.
If one of these intermediate regions is damaged — as can occur with conditions such as Alzheimer’s disease or traumatic brain injury — the translation process can fail, and new long-term memories may not be formed. This explains why people with hippocampal damage often retain older memories formed before the injury but struggle to create new lasting memories.
Mimicking the brain’s translation process
Using neural recordings gathered from animals and then from human patients, Song and Berger devised a model that accurately reproduces how the brain translates short-term signals into the neural patterns associated with long-term storage. The prosthetic system is intended to bypass a damaged hippocampal section by taking the signal produced in an earlier region, transforming it according to the model, and delivering the appropriately translated pattern to the downstream region as if the damaged tissue had performed the translation itself.
The researchers emphasize that the device does not “read” memories in the sense of decoding specific thoughts or experiences. Instead, it models and recreates the electrical transformation that normally takes place between hippocampal subregions — a form of neural translation rather than semantic interpretation. As Berger put it, the approach is analogous to translating from Spanish to French without understanding the content of either language: the system reproduces the mapping between patterns, not the meaning they carry.
Testing and results
To evaluate the model, clinicians who had implanted electrodes in patients’ hippocampi to treat chronic seizures recorded the electrical activity associated with memory formation in two hippocampal regions. Those recorded signals were provided to the modeling team, who trained and validated the algorithm by comparing how it transformed signals from the first region into the patterns observed in the second region.
Across hundreds of trials in nine human patients, the algorithm predicted the downstream hippocampal signals with roughly 90 percent accuracy. That level of predictive performance indicates the model can replicate the neural transformations that normally occur during memory encoding, supporting the possibility of a device that replaces or augments the function of a damaged hippocampal region.
The next experimental step planned by the team is to deliver the model-generated, translated signals back into a patient’s brain at the downstream site. The goal is to bypass the damaged hippocampal region and thus restore the neural conditions needed to form accurate long-term memories.
Funding: The research, which may ultimately assist wounded service members and others with memory impairment, was supported by DARPA (project N66001-14-C-4016 to Wake Forest Baptist Medical Center) and USC.
Source: Robert Perkins – USC
Image credit: Image adapted from a USC press release
Original research: The work was presented at the 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society in Milan.