Summary: Researchers have achieved a major advance in neural prosthetics by using a new memory decoding model (MDM) to successfully recall specific human memories. Building on prior work, the team used surgically implanted electrodes in the hippocampus to decode neural activity patterns and apply targeted electrical stimulation to enhance memory encoding and retrieval.
In a clinical study of 14 adults with epilepsy, MDM-guided stimulation produced meaningful improvements in memory performance, especially for participants with impaired memory. The results point toward potential future therapies for memory loss caused by Alzheimer’s disease, stroke, or head injury and represent a significant step toward implantable devices that could restore lost memory functions and support greater independence.
Key Facts:
- The neural prosthetic uses a memory decoding model to identify hippocampal activity patterns associated with specific memories and then applies targeted stimulation to enhance recall.
- Overall, stimulation altered memory performance in roughly 22% of tested patient-and-category combinations, with increases in memory occurring at nearly a 2-to-1 ratio over decreases. Participants with impaired memory who received bilateral stimulation showed changes in nearly 38% of combinations, with increases outnumbering decreases by more than 4-to-1.
- Funded by the U.S. Defense Advanced Research Projects Agency (DARPA), this work advances avenues for treating memory disorders and developing neural prosthetics to aid daily living.
Source: Wake Forest University
A collaborative team from Wake Forest University School of Medicine and the University of Southern California (USC) reports the first successful use of a hippocampal neural prosthetic to recall specific memories in human patients.
These findings are published in Frontiers in Computational Neuroscience and extend the team’s earlier 2018 work led by Robert Hampson, Ph.D., which demonstrated a prosthetic system that used a person’s own neural activity patterns to support memory encoding and recall.
Previous prosthetic systems were based on a multi-input multi-output (MIMO) nonlinear model that influenced firing patterns across multiple hippocampal neurons. In this study, the researchers developed a memory decoding model (MDM) to map the spatiotemporal firing of hippocampal neural ensembles that encode specific items of information (for example, “I turned off the stove” or “Where are my keys?”).
Using intracranial recordings from implanted electrodes, the MDM decodes which patterns of neural activity correspond to particular pieces of information. Those decoded patterns were then translated into stimulation codes delivered to CA1 and CA3 regions of the hippocampus during the memory encoding phase, with the goal of reinforcing the neural representation of targeted content.
“We demonstrate an innovative approach to neurostimulation that is not just general enhancement but is tailored to specific, behaviorally relevant information for each person,” said Brent Roeder, Ph.D., a research fellow in translational neuroscience at Wake Forest University School of Medicine and the study’s corresponding author.
The clinical trial enrolled 14 adults with epilepsy who were undergoing diagnostic brain mapping with surgically implanted electrodes to locate seizure foci. Surgical procedures, monitoring, and cognitive testing were performed at Atrium Health Wake Forest Baptist Medical Center, Keck Hospital of USC, and Rancho Los Amigos National Rehabilitation Center.
Researchers applied MDM-derived electrical stimulation during visual recognition memory tasks to test whether stimulation during encoding would improve later recognition. Across all participants and stimulus categories, stimulated trials produced significant changes in memory performance in about 22% of patient-and-category combinations. These changes included both improvements and declines, with improvements occurring nearly twice as often as declines.
Notably, among participants with impaired memory who received bilateral hippocampal stimulation, almost 38% of patient-and-category combinations showed significant changes, and improvements outnumbered declines by more than four to one. The results indicate that the effectiveness of MDM-based stimulation depended on baseline memory function and whether stimulation was unilateral or bilateral, with the greatest gains seen in impaired subjects receiving bilateral stimulation.
“Our aim is to develop interventions that restore memory functions lost to conditions like Alzheimer’s disease, stroke, or traumatic brain injury,” Roeder said. “We observed the most pronounced benefits in participants with existing memory impairment, suggesting this approach could be clinically valuable.”
The team envisions refining this technology to support everyday tasks—helping people remember whether they’ve taken medications, turned off appliances, or locked doors—so they can live more independently. While additional research and clinical testing are required, MDM-guided stimulation shows promise as a method to selectively modify memory encoding and recall.
This human study builds on more than two decades of preclinical research on hippocampal memory codes led by Sam Deadwyler, Ph.D., and collaborators including Robert Hampson, Theodore Berger, Ph.D., and Dong Song, Ph.D. Earlier animal studies applied similar stimulation approaches with the MIMO system developed at USC and demonstrated restoration and facilitation of memory in animal models.
Funding: U.S. Defense Advanced Research Projects Agency (DARPA).
About this neurotech research news
Author: Myra Wright
Source: Wake Forest University
Contact: Myra Wright – Wake Forest University
Image: Image credited to Neuroscience News
Original Research: Open access. “Developing a hippocampal neural prosthetic to facilitate human memory encoding and recall of stimulus features and categories” by Robert Hampson et al., Frontiers in Computational Neuroscience.
Abstract
Developing a hippocampal neural prosthetic to facilitate human memory encoding and recall of stimulus features and categories
Objective: This study demonstrates the first effective use of static, content-specific neural stimulation patterns derived from a subject’s own hippocampal spatiotemporal codes to influence memory encoding and recall.
Approach: The investigators constructed a memory decoding model (MDM) that characterizes how hippocampal CA3 and CA1 neural ensembles encode specific memory items through spatiotemporal firing. The MDM generates stimulation patterns applied to CA1 and CA3 during the encoding (sample) phase of a delayed match-to-sample (DMS) short-term memory task in human subjects.
Main results: MDM electrical stimulation delivered to CA1 and CA3 during the sample phase improved recognition of DMS task images during a delayed recognition assessment that also included control images. Across all subjects, stimulated trials produced significant performance changes in 22.4% of patient-and-category combinations. Overall, increases in memory performance nearly doubled the number of decreases. In patients with impaired memory who received bilateral stimulation, significant changes occurred in over 37.9% of combinations, with increases outweighing decreases by more than four to one. The effect depended on baseline memory function and whether stimulation was unilateral or bilateral, with most improvements seen in impaired subjects who received bilateral stimulation.
Significance: These findings show that it is possible to facilitate encoding of specific memory content in patients with impaired memory, supporting the development of future implantable neural prosthetics aimed at improving human memory.