Penn’s Restoring Active Memory Project Releases Extensive Human Brain Dataset

A University of Pennsylvania project aimed toward building a device that helps improve memory for patients with memory disorders has reached an important milestone.

Two years into the DARPA-funded Restoring Active Memory or RAM program, lead researcher Daniel Rizzuto, director of cognitive neuromodulation, and Michael Kahana, Penn psychology professor and RAM principal investigator, along with colleagues, have enrolled more than 200 patients and collected more than 1,000 hours of data from patients performing memory tasks. They have now released the largest human intracranial brain recording and stimulation dataset to date, and it’s available for public use, for free.

“This project is really unprecedented in its scope and scale. It’s very much a team effort,” Rizzuto said. “We’ve brought together an amazing group of clinicians, scientists and engineers at 10 organizations across the country to develop a revolutionary new technology to improve memory. Along the way we plan to release all of the project’s data to the public so that other teams might benefit as well.”

The end-point of RAM is to develop a fully implantable device that can electrically stimulate the brain to improve memory function. The program’s immediate goal is to deliver new treatments for those who have experienced a traumatic brain injury, such as veterans returning from combat. In the long term, such therapies could help patients with a broad range of ailments, from Alzheimer’s to dementia.

However, the research isn’t quite there yet.

Implanting such a device requires brain surgery, so initially the RAM team had to better grasp how each part of the brain functions.

“Understanding the basic neurophysiology of human memory in each patient allows us to then intervene using brain stimulation to enhance memory performance,” Kahana said.

Given the early stage of this technology’s development, the scientists opted to work with epilepsy patients who already had electrodes implanted into their brains as part of their routine clinical care. Eight clinical centers across the country enrolled participants into a standard protocol that included cognitive testing sessions during which the Penn team gathered high-precision recordings and used stimulation to elucidate connections among the brain’s memory centers and to refine mechanisms for decoding and encoding neural activity. Participating medical centers included Columbia, Emory and Thomas Jefferson universities; the Dartmouth School of Medicine; the Mayo Clinic; the University of Texas Southwestern; the National Institute of Neurological Disorders and Stroke; and the Hospital of the University of Pennsylvania.

Cognitive tasks, chosen for their importance in carrying out activities of daily living, included verbal free recall, where patients were asked to remember a set of words, perform a few simple arithmetic problems to distract them from the list, then recall as many of those initial words as possible. There was a paired-associate learning task, as well as a spatial navigation task during which subjects moved around a computer-generated environment, similar to a computer game.

“We’re getting lots of data from these patients because we must account for the many different aspects of memory as we build our model,” Rizzuto said. “Every patient’s brain is different, and we need to develop a patient-specific model for how the memory system works to optimize the delivery of stimulation. That’s our approach.”

The recently released dataset includes information from 700 sessions, and for every patient, intracranial recording files from 100 to 200 electrode channels, neuro-anatomical information indicating the location of each electrode, precise records of patient behavior and the experimental design documents. To receive the raw dataset, interested researchers may request access through the RAM website.

For their part, the RAM researchers say they will continue using the data they’ve gathered to refine their algorithms and identify the optimal stimulation location for the implantable device.

“We’re essentially developing a recipe for how to record and stimulate the brain in order to restore memory. And we’ve got 200 patients’ worth of data that we’re using to refine that recipe,” Rizzuto said. “At the end of Penn’s RAM effort, which we anticipate will be about 20 months from now, we hope to have that recipe finalized.”

It is one more step toward an implantable device that’s not too far off in the future.