Small, fiber-based endoscopes have improved our ability to image deep within the human body. Current fiber-based devices consist of bundles of single mode fibers (SMF) in which individual fibers represent single pixels in the transmitted image. Alternatively, a single SMF can be mechanically scanned to create an image. The use of a multi-mode fiber (MMF) in imaging is desirable because of its potential to image with a higher resolution and a smaller diameter than SMF bundles without mechanical scanning, thus allowing access to more remote parts of the body. However, image transmission through MMFs is limited by mode coupling, which scrambles the image information. In this paper, Ploschner, et al. expand upon recent works to unscramble the transmission of light through MMFs and introduce a GPU toolbox to make these techniques faster and more accessible to researchers.

The basis of unscrambling information through a MMF is via the measurement of a transformation matrix through the fiber. This transformation matrix can then be combined with wavefront shaping to control light propagation through the fiber. In this paper, the authors describe a new GPU-based, open-source toolbox for transformation matrix measurement and light control through a MMF. The GPU enables maximum frame-rate control of the wavefront-modulating spatial light modulator (SLM) for optimal measurement time and real-time control of a transmitted intensity pattern. Furthermore, the paper describes system improvements to eliminate interference effects in intensity pattern creation.

To measure the transformation matrix through the fiber, a diffraction limited spot is scanned across the input facet of the MMF. For discrete locations of the input facet, the transmitted complex field of the emitted speckle pattern is measured through phase-shift interferometry with a SMF providing a reference. This information, referred to as the transformation matrix, provides a complete characterization of light propagation through the fiber. The toolbox stores this in the GPU memory and uses it for real-time intensity pattern generation with the transmitted light.

The transformation matrix allows for the creation of arbitrary image patterns at the MMF output. This can be done through the superposition of the calculated phase masks to create the individual focus points of the desired pattern. However, superposition introduces deleterious interference effects between focus points. To overcome this, the authors utilize an acousto-optic deflector (AOD) to modulate the incidence angle onto the SLM to selectively transmit only a single phase-mask at a time. The phase masks remain superposed on the SLM, but each phase mask is encoded with a unique grating to match an incidence angle. Thus, the pattern focal points are displayed individually in a time-sequential way and do not interfere. With the AOD all of the pattern focus points are transmitted within the pattern refresh rate of the GPU (50 Hz).

The authors also demonstrate the real-time pattern generation capabilities of the GPU toolbox. An interactive routine provides user control of an intensity pattern created through the fiber. The pattern displayed is a rotating cube, which is calculated and refreshed at 50 Hz. Furthermore, to provide support for those interested in using the toolbox, the appendix includes a detailed description of how to set up and use the LabView interface for the GPU toolbox.

The introduction of this open-source GPU toolbox, with a convenient high-level user interface, makes MMF implementation in endoscopic applications more accessible to interested researchers. The GPU allows for SLM control at maximum frame rate and real-time calculation and updating of the SLM for arbitrary intensity pattern creation. The ideas and toolbox presented in this paper will accelerate the development of MMF endoscopes toward minimally invasive, high resolution imaging deep within the body.

---

---

 Add Comment

You must log in to add comments.