Neuroscientists have 'hacked' the brain's memory pathways with a new prosthetic system that could help improve the recollection of specific memories.

The deep brain technology can't remind you whether you've turned off the oven just yet, but research led by Wake Forest University (WFU) and the University of Southern California (USC) has now shown it can sometimes help people recall particular images with greater ease.

Electrical and magnetic brain stimulation have recently emerged as promising new ways to give overall cognitive performance a boost, but until this recent breakthrough the idea of a brain 'zap' to recollect discrete details was the work of science fiction.

"Here, we not only highlight an innovative technique for neurostimulation to enhance memory, but we also demonstrate that stimulating memory isn't just limited to a general approach but can also be applied to specific information that is critical to a person," explains neuroscientist Brent Roeder from WFU School of Medicine.

The experiments were conducted among 14 adults with epilepsy, who were being fitted with brain electrodes to locate the epicenters of their seizures.

Roeder and his colleagues have been working with similar cohorts for years now, as these implants also provide a platform from which to study electrical brain stimulation.

In 2018, the team tested the neural implants by asking them to 'write codes' of information into the hippocampus – the place where memories are said to be 'stored' in the brain.

These electrical codes, the researchers explain, were designed to emulate a pattern of neural activity associated with remembering information.

Now, the team has gotten even more specific. They have used a computer model that essentially watches a person's brain activity to try and figure out what patterns align with the memory of specific images.

When participants took part in a visual memory test, this model generated stimulation patterns for each individual, and these patterns were tied to memorizing images of animals, buildings, plants, tools, and vehicles.

When those same neural 'codes' were artificially fed back into the hippocampus, participants in the study were better at matching previously observed images from memory in about 22 percent of cases.

That is a relatively low accuracy, however results jumped to nearly 38 percent when researchers stimulated both hemispheres of the brain and focused only on participants with impaired memory.

"Our goal is to create an intervention that can restore memory function that's lost because of Alzheimer's disease, stroke or head injury," says Roeder.

"We found the most pronounced change occurred in people who had impaired memory."

The team of researchers say that their findings are clearly indicative that this form of deep brain stimulation "has the potential to be used to significantly modify memory". Yet multiple challenges remain in the way of that future.

Finding 'static codes' for individual memories is a lofty goal that requires the model at hand to be further refined and tested. Even when a code works in boosting memory, as in the current experiment, it's unclear what component of an image category those neural patterns are actually encoding.

"An image of a house with a tree in it could be intended to have the house be the main focus of the image by researchers," the authors of the current study explain, "but the position of the tree could interfere with this intent."

Focus may also differ from person to person, and over time.

The researchers plan to keep tinkering with their memory model to figure out how more basic information is coded and retrieved in the hippocampus. They also want to test whether the memory patterns of one individual can be used to stimulate memory in another.

If that turns out to be possible, then it suggests that the prosthetic memory system is 'writing' a memory code into the brain, as opposed to 'reinforcing' a code that is already present.

"Each of these questions will move the research forward to the point of developing a memory prosthetic operating on general features of memory encoding that are common across patients, yet specific enough to facilitate retention of specific memory content," the team concludes.

The study was published in Frontiers in Computational Neuroscience.