In the 1987 movie Innerspace, Dennis Quaid—riding in a special craft shrunk by experimental technology—gets injected into Martin Short's body. There, he's treated to a very up-close-and-personal look at the body, from the optic nerve to the blood vessels. While this was just a sci-fi comedy, scientists have recently created a device that brings it one step closer to reality: a camera the size of a grain of salt.
A basic camera consists of two parts: a lens that takes in incoming light, and a sensor that records it to produce an image. Unfortunately, lens and sensor technology haven't exactly kept pace with one another. While smaller sensors have allowed devices like smartphones to serve as pretty advanced cameras, lenses have been pretty much the same since they were invented.
Conventional lenses are made of glass or plastic. They're shaped with a curve, in order to bend incoming light before it reaches the sensor. Changing this curve changes how the light interacts with the sensor, which allows for macro shots, telephoto shots, and “fisheye” effects. Their construction means that there's a limit to their miniaturization—a lens can only be so small before it stops doing what you want it to do.
The lenses for this new, tiny camera rely on a metasurface. The lens is actually a compound lens, made up of many tiny nano-antennas placed on a super-thin surface. Because they're made of such tiny structures, they can manipulate light in new and interesting ways. Depending on the structure, shape, and proximity of the nano-antennas, they can bend or reflect light as needed.
While this is a pretty advanced technology, it's not really new. Scientists were already able to make lenses like this, but they weren't really usable. Their images were often blurry, distorted, or had a field of view that was too narrow to be useful for much. Machine learning is changing all of that.
By pairing the images captured by the tiny camera with a machine learning algorithm, researchers were able to capture images that were comparable in quality to a compound lens many times bigger than the miniature metasurface camera.
For a long time, getting a good look at the inside of the body meant cutting into it. Some systems, like the digestive tract, can be observed using cameras that often require that the patient be sedated. This limits the areas that doctors and researchers were able to see. Even technology like computed tomography (CT) scans and magnetic resonance imaging (MRI) are only able to do so much and rely on injectable contrast material to get a decent look at blood vessels. Examining the brain, in particular, has always been next to impossible.
The brain is extremely well-protected, and for good reason. While this is definitely better than the alternative, it does severely limit the ways that we can access the brain to examine it. Its complexity and abundance of blood vessels also make interacting with the brain extremely challenging—it doesn't take much to cause severe neurological problems or even death.
This combination of camera and algorithm could represent an incredible advance in neurology. The ability to capture images of the brain could not only be used in experiments to further the science of neurology, but it could also be used as a diagnostic or monitoring tool to help people with specific problems. Scientists could get full-color, focused images of structures in the brain without needing to open the skull or potentially damage any of its tissue. They could directly observe some of the mechanisms that may be responsible for the development of conditions like stroke, Alzheimer's disease, and intracranial hypertension.
Nearly one in six people in the world suffer from some form of neurological condition. This camera could conceivably change lives for hundreds of millions of patients.
While the idea and technology for producing a miniaturized compound lens isn't new, the combination of lens and machine learning is. With the aid of a properly trained algorithm, these tiny, sand-grain-sized lenses can infiltrate areas that were previously unseen and return images that could improve the diagnosis and treatment of neurological conditions.
