Real-time adaptive brain control: Combining a BCI with DBS to treat Parkinson’s

We have been following the advances in deep brain stimulation (DBS) pretty closely here at ET. Just the other week, we explained the motivations behind treatments for Parkinson’s disease with DBS electrodes placed in the subthalamic nucleus, and hinted at a couple of ways that they might be improved. Perhaps the biggest advance in DBS systems since they were first developed has just been reported by a group of researchers working in the Movement Disorders Group at Oxford University. By recording activity with what is essentially a brain-computer interface (BCI), the researchers were able to close the stimulation loop with direct feedback from the subthalamic nucleus. The ability to incorporate meaningful data from the implant user’s brain into the moment-to-moment control of the stimulation puts the one-size-fits-all DBS system on the fast track to obsolescence — and ushers in the new era of personalized implants.
Once activated, most DBS systems generate continuous pulses, typically at a frequency above 100Hz. Always-on stimulation fatigues not only the batteries, but more importantly the brain itself. The symptoms of Parkinson’s disease fluctuate continuously and any form of smart control needs to be fast to be effective. Previous research has demonstrated that the spike output of motor neurons in the cortex of a monkey could be used as feedback to provide better results than continuous stimulation. Directly implementing these results in humans calls for a bit more caution though — blunt exploratory implantation of multiple brain areas comes at a price. The Oxford group, led by Peter Brown, realized that the stimulation electrodes themselves can be used to record local field potentials. These signals are not the signatures of single cells, but can be thought of instead as more diffuse summary reports of larger groups of cells.

In order to use these averaged potentials for control, some criteria is needed to pass judgement on their desirability. Studies in Parkinson’s patients have consistently shown that subthalamic nucleus activity in the beta frequency band (3-30Hz) correlates with motor impairment. Brown’s group was able to build a control system that was able to filter and capture activity in this band, and then use it to control the stimulation current. The device was tested in eight patients and provided significant improvements over both continuous and random stimulation conditions.
This proof-of-principle demonstration provides a tantalizing glimpse of what lies ahead. For now, the control hardware is a bulky external system, although work on miniaturizing its footprint to fit inside the skull is ongoing. Once that happens, we have the essentials of two closely-related kinds of devices that are also under development in several labs. One is a device to police undesirable activity, like for example, seizures. In fact an important study in this area just appeared in the journal Neurosurgery. The other, more intriguing use is for what is commonly referred to as a “memory implant.”

DBS electrodes, reaching down into a human’s brain

There is really no substantial, convincing, evidence yet that memory implants are practical. The main problem is that we don’t know how to interpret what little brain activity we can manage to record. What I suspect we will find going forward is that, rather than attempting to build memory implants from scratch that access high-level areas of the brain (like the hippocampus), they will instead directly evolve from the on-board learning algorithms, and storage, that are built into motor control implants like those described here.
The pace of advance of these kinds of implants will be driven to a large degree by the pace at which they are made open. Interoperability between components is essential, as well as keeping the user in the loop to provide direct input to device designers. A huge boost in this area was just provided by Medtronic, the leading implant manufacturer, when it finally caved to overwhelming user demand for access to critical data generated by their implants. The irony of users having better feedback from an inexpensive Fitbit than from their $30,000 implant was not lost on vocal advocate, user, and TED Talk giver, Hugo Campos. Others were quick to jump on Medtronic CEO, Omar Ischrak when he tweeted how much he enjoyed access to his Runtastic data, yet at the same time denying feedback to users of his own products.
Opening implant data to users is harmless enough, although giving them unbridled control over critical parameters certainly has its risks. The larger community has already realized what the medical community has been slow to accept — the benefits of providing users with control over their implants far outweighs these risks, and that fact can no longer be ignored. Increasingly, we interact with our external environment through our personal electronic devices — that same power, and ease of interaction, for our inner space will be sought, demanded, and granted.