Shrinking optical endoscopes

Imaging hidden structures in the body, either for medical diagnosis or treatment, is a challenge. An endoscope is a device often used to look inside the body and usually comprises parts such as an optical fiber to bring the light and some optical components, like lenses, to direct it. Additionally, some mechanical actuators are used to move the fiber tip to form an image by collecting scattered light information from different spatial locations. The diameter of an endoscope integrating all these components can reach a few millimeters. However, for certain applications, such as looking inside the ear, even this is too large. Thus, there is a need for miniaturized endoscopes that provide high-quality images.

Two principal approaches have been used to make minimally invasive, i.e., small-diameter, endomicroscopes. The first is based on a bundle of small-diameter optical fibers (up to 100,000) that deliver illumination and at the same time collect light from tissues. With this method, the distance between the fibers limits the resolution, and small structures cannot be seen.¹ The other approach uses a small-diameter fiber and a lens to focus the light on one point, and then the point is moved by actuators propelling the whole system. This results in a bulky system.¹ To get around these issues, we have been working on an endoscope that integrates a single multimode fiber. This endoscope is as thin as the fiber itself and provides high-resolution images of small structures without any mechanical actuators.

Figure 1. Schematic of the multimode-fiber-based endoscope: a readout beam picks a ‘specific’ wavefront (assigned phase) on an SLM (spatial light modulator) and projects it on the fiber's input to create a clean focus spot at the fiber output. The focus spot is moved by changing the phase pattern.

Using multimode fibers for both light delivery and collection in an endoscope is very recent.² A multimode fiber is an optical fiber that has a large core and a large acceptance angle. However, these fibers are limited by their multimodal property, meaning that when light is injected in them, it gets scrambled. Thus no image or information can be carried through them without being distorted. Our approach compensates for the distortions induced by a multimode fiber by controlling light propagation through it with a spatial light modulator, or SLM (see Figure 1). This way, we can generate a focus spot and move it without using a lens or an actuator. Thanks to this method, multimode fibers provide lensless imaging with submicrometer resolution over the diameter of the fiber and therefore offer a means to further miniaturize endoscopes.^2–4

Figure 2. Left: Output of a multimode fiber: light is scrambled and no direct imaging is possible. Right: Grid of foci generated at the fiber output using the method called digital phase conjugation. w₀: Spot size at the focus.

Figure 3. Left: Fluorescence images of stained cochlear hair cells for the diagnosis of hearing loss. Right: Fluorescence images of stained neural cells and dendrites (scale bars: 10μm).

We have experimentally demonstrated a record ultrathin endoscope of diameter 0.5mm, capable of imaging a field of view of 200μm with submicrometer resolution. The image is formed without moving parts and by digital means only. A multimode fiber supports many optical modes, and the output wavefront is typically a speckle pattern: see Figure 2 (left).

In our endoscope, a phase-only SLM produces a ‘specific’ wavefront to the input of the multimode fiber to create a clean focal spot at the output of the fiber: see Figure 1. To find the specific wavefront, we first ‘learn’ the fiber characteristics by using a focus spot placed at the fiber output. Next we record the transmitted wavefront at the input. We then reverse this wavefront so that a second pass through the fiber from the input to the output will produce a clean focus spot, at the same position as the initial one. This process is called optical phase conjugation.³

In our setup we realize phase conjugation through a combination of SLM and camera, making the treatment all-digital, without mechanical actuators. We can thus store many phase patterns to focus on the full range of desired positions. The phase patterns are then projected sequentially onto the SLM, and the readout beam picks up the phase information to give just the right wavefront to focus the beam at the intended location at the fiber output. This way, a grid of thousands of focus spots is generated by changing the phase patterns displayed on the SLM: see Figure 2 (right) for an example of a grid generated over a 60μm field of view.

To perform imaging, at each focal spot the fluorescence generated by the sample under test is collected by the same large-aperture multimode fiber, which is guided to the fiber input and detected. By using an even a larger fiber core, supporting around 80,000 modes, we can generate submicrometer spots that provide high-resolution images. With this system we were able to image cochlear hair cells and neural cells: see Figure 3.² The quality of the acquired images assists effective diagnosis that depends on cellular phenotype.

In summary, by controlling the light that we inject into the multimode fiber, we can generate and scan a clean focus spot. Thus, a single multimode fiber is used as a thin, rigid, high-resolution endoscope that is lens and actuator free. Our next step is to implement the system in a portable device and perform imaging in vivo.