AI-enhanced brain sensor tracks poorly understood chemistry

Researchers have developed a device that can simultaneously measure six markers of brain health. The sensor, which is inserted through the skull into the brain, can pull off this feat thanks to an artificial intelligence (AI) system that pieces apart the six signals in real time (ACS Sens. 2024, DOI: 10.1021/acssensors.4c02126).

Being able to continuously monitor biomarkers in patients with traumatic brain injury could improve outcomes by catching swelling or bleeding early enough for doctors to intervene. But most existing devices measure just one marker at a time. They also tend to be made with metal, so they can’t easily be used in combination with magnetic resonance imaging.

The new device is just 2.5 mm across and relies on metal-free fiber optics to measure physical and chemical properties of the brain: temperature, pH, and the concentrations of sodium ions, calcium ions, glucose, and dissolved oxygen. Those biomarkers were chosen based on previous studies of how they change in the cerebrospinal fluid (CSF) in the spinal cord, says Ali Yetisen, a chemical engineer at Imperial College London, who led the study along with Nan Jiang of Sichuan University. While researchers believe the measurements should give them insight into the energy metabolism of the brain, these markers have not been extensively studied, because previous equipment has not been able to measure them in real time and simultaneously, Yetisen says.

The sensor consists of seven optical fibers—one for each biomarker plus a spare—each coated with a commercially available small molecule or enzyme that fluoresces when it interacts with the marker it’s looking for. The fibers are inserted into the brain using a soft, flexible catheter designed to minimize damage to brain tissue.

Each fluorescing molecule is encapsulated in a polymer chosen to work with the property it measures. For instance, the dissolved oxygen probe is coated in polydimethylsiloxane, which is permeable by oxygen but not water. For the ion probes, the team used a poly(ethylene glycol) diacrylate–acrylamide hydrogel, which admits water but keeps the sensor’s fluorophore stable, says Yubing Hu, a research associate at Imperial College London who worked on the device.

A laser that can emit three different wavelengths is shot through the fiber bundle to trigger the fluorescence. And finally the fluorescent signal bounces back up the fibers into a light sensor.

An issue with measuring so many signals simultaneously is that some of them overlap in the researchers’ spectra or get lost in background noise. To get around that, the researchers trained an AI system called a neural network. The system learned from data the device gathered from three sources: a lamb brain immersed in artificial CSF, a laboratory solution, and CSF taken from patients at a local hospital. The AI essentially makes the sensor more powerful because it can pick up signals that are otherwise hard to spot.

The AI can also identify characteristics of the signals, such as their intensity. It can predict changes in the markers in real time, so if it finds, for instance, an increase in sodium levels that might indicate a problem, it can alert doctors.

The device will need to be tested in live animals and then in clinical trials in humans before it can be cleared for clinical use, a process that Yetisen says could take 5–10 years.

“This looks like a clever strategy for multimodal sensing,” says John Rogers, a chemist at Northwestern University who was not involved in the research. “It will be interesting to determine in future work whether this approach can be used for long-term monitoring.”