An external auxiliary brain made from software, a chip or lab-grown tissue could take over computation lost when a stroke destroys the cortex, a neurologist proposes

An external auxiliary brain made from software, a chip or lab-grown tissue could take over computation lost when a stroke destroys the cortex, a neurologist proposes

On October 21, Frontiers in Human Neuroscience published a theoretical paper by Mikhail Serruya outlining the Auxilium project and a plan for experiments in humans.

A surgeon identified in the paper as Dr. G. had a stroke. Rehabilitation restored strength to his arm, but his speech remained impaired: he struggles to produce words and cannot put them together grammatically.

Speech interfaces read signals from intact cortex in people who have lost the ability to speak because of paralysis. In Dr. G.’s case, the tissue that generated speech had died. Such interfaces, Serruya writes, “cannot read what is no longer there.” In the brain interface literature, Serruya identifies an assumption he calls the “bandwidth paradigm”: more channels mean greater capabilities. In a 2019 paper, Neuralink attributed the limited clinical adoption of brain interfaces partly to their “modest number of channels.”

Serruya challenges this assumption with a comparison to USB ports: adding ports gives a computer more inputs while leaving its processor unchanged. In his view, what distinguishes humans from chimpanzees is the number of loops connecting the cortex and subcortical structures. Dr. G. has lost part of the processor itself.

Auxilium is a system outside the skull that operates alongside the brain and exchanges signals with the surviving cortex in both directions. “Virtual white matter” restores connectivity by recording activity at one site and stimulating another, replacing the bundles of fibres that connect brain regions. An external module modelled on cortical circuitry, or lab-grown tissue, takes on the computation. Serruya proposes building the module with corresponding loops, because without them the added cortex could encounter a bottleneck.

Scarring after a stroke and the risks of operating on an injured brain may limit the feasibility of implanting tissue inside the skull. In rats with stroke, transplanted human neural progenitor cells survived at four weeks only when accompanied by an enzyme that breaks down molecules in scar tissue. Serruya suggests the abdomen as one possible alternative: it offers more space and a good blood supply. “Living electrodes,” bundles of lab-grown neuronal extensions used in place of wires, would be woven into a cable running beneath the skin from the brain to the abdomen. There, a lab-grown “third hemisphere,” with its own cortex and subcortical structures, would be connected to blood vessels in the omentum, a fatty fold of the peritoneum. Surgeons have already taken tissue from the abdomen to the brain: a 1999 case report described a woman who had suffered a stroke and whose speech and mobility improved, according to the authors, after her omentum was transferred to her brain. She was followed for 13 years.

Serruya proposes testing the idea in people who already have depth electrodes implanted to treat epilepsy. Modules from five classes, ranging from simple signal relays to living neuronal cultures, would be connected to the brain with latency and bandwidth held equal, so that any differences would reflect the module’s design alone. In the video game Ms. Pac-Man, participants would receive directional cues through brain stimulation, generated either by a simple program or by a living neuronal culture. If relaying signals between regions alone improves performance, connectivity would be sufficient; if improvement requires a module with activity of its own, its computation would be needed. The plan has six stages of increasing risk. People with stroke would enter at the third stage. At the final stage, involving tissue in the abdomen, that tissue would have to outperform synthetic modules by a margin large enough to justify the surgical risk.

To Serruya’s knowledge, no one has conducted such a comparison under matched conditions. Serruya founded Neurodelphus, a company working on brain interfaces and computation using living tissue, and disclosed a financial interest in the company in the paper.