This PDF file contains the front matter associated with SPIE Proceedings Volume 10886, including the Title Page, Copyright information, Table of Contents, Author and Conference Committee lists.

Normal development of the visual system in infants relies on clear images being projected onto the retina, which can be disrupted by lens opacity caused by congenital cataract. This disruption, if uncorrected in early life, results in amblyopia (permanently decreased vision even after removal of the cataract). Doctors are able to prevent amblyopia by removing the cataract during the first several weeks of life, but this surgery risks a host of complications which can be equally visually disabling. Here, we investigated the feasibility of focusing light noninvasively through highly scattering cataractous lenses to stimulate the retina, thereby preventing amblyopia. This approach would allow the cataractous lens removal surgery to be delayed and hence greatly reduce the risk of complications from early surgery. Employing a wavefront shaping technique named time-reversed ultrasonically encoded (TRUE) optical focusing in reflection mode, we focused 532 nm light through a highly scattering ex vivo adult human cataractous lens of 112 mean free path thick. This work demonstrates a potential clinical application of wavefront shaping techniques.

Surgery with fs-laser in the posterior part of the eye could be useful for separation of tractional epiretinal membrane and vitreous floaters treatment. However, focus degradation occurs near the retina due to induced aberrations by cornea and lens. To overcome this issue, adaptive optics with wavefront sensor and wavefront modulator can be utilized. We demonstrate an alternative concept for image guided femto second laser (fs-laser) surgery in the posterior eye with wavefront sensorless adaptive optics (WFSLAO). Our laboratory setup consists of an 800 nm fs-laser and a superluminescent diode (SLD) with 897.2 nm central wavelength. The SLD is used for optical coherence tomography (OCT) whereby the light for the OCT sample arm and the fs-laser share the same optical path which contains a deformable mirror, scanner and focusing optics. Energy calibrated photodiodes are used to measure the threshold energy for a laser induced optical breakdown inside a water filled chamber that acts as simple eye model. OCT image based metrics were used to determine a set of Zernike polynomials that describe a near optimal deformable mirror state. Such a mirror state improved OCT resolution and at the same time lowered the required fs-laser energy for a laser induced optical breakdown inside the eye model substantially.

Neuroscience research related to functionality, connectivity and metabolism of neuronal circuits, individual neuronal cells and sub-cellular structures, nowadays, experiences a burgeoning need to develop techniques for the detailed investigation inside the complexity of the living matter. Particularly, high-resolution observations combined with an extended depth of penetration in tissue represents an ongoing challenge. Holographic control of light propagation in complex media opens a promising way to overcome this technological barrier via exploiting multimode fibres as hair-thin, minimally-invasive endoscopes. This concept allows for more than one order of magnitude reduction of the instrument’s footprint and a significant enhancement of imaging resolution, compared with current minimally invasive endoscopes. Here, we demonstrate a compact and high-speed system for fluorescent imaging at the tip of a fibre. The instrument’s performance reaches micron-scale resolution across a field of view 50 micrometres, yielding 7-kilopixel image information at a rate of 3.5 frames per second. The resolution limit is dictated only by the numerical aperture of the fibre probe, and the contrast/pureness of the focal points, utilised for raster-scanning regime, approach the theoretical limits for phase-only holographic wavefront shaping. The achieved performance allowed for in-vivo observations of neuronal somata and processes, residing deep inside the visual cortex and hippocampus of an animal model with minimal damage to the tissue surrounding the fibre penetration area. We believe that this demonstration represents an important step towards implementations of various advanced forms of imaging through multimode fibre based endoscopes to address numerous key challenges in neuroscience.

Scattering is a major obstacle on the way of imaging deeper than a few mean-free-paths through bone. The high density of mineralization and collagen fibers deposition make bone a very inhomogeneous tissue that produces severe scattering. Although long wavelength excitation extends the mean-free-path for multi-photon microscopy, however imaging more than 150 microns through bone suffers from loss of resolution and intensity. We previously simulated the wavefront distortions caused by bone using phase accumulation ray tracing (PART) method. Our findings show that some low-order optical aberrations can be corrected using traditional adaptive optics systems such as a deformable mirror, however, a significant amount of high order aberrations are remaining, which require a secondary correction method to restore the point spread function at depth. In this work, we use a high-speed binary wavefront correction method using a digital light processor (DLP) to correct the wavefront in a hostile environment such as bone. We use the PART method to produce an initial estimate of the wavefront, and use a genetic algorithm to evolve it to an optimum using maximum intensity metric. The binary wavefront correction produces a factor of 21 enhancement and the initialization using PART method increases the enhancement 2.5 times.

Zebrafish are an important vertebrate model used to view the mechanisms underlying seizure disorders. Due to their relatively small size and transparency, larval zebrafish are an excellent model through which to view the occurence of seizure-like neural activity in vivo using light sheet fluorescence microscopy (LSFM). Although LSFM possesses good optical sectioning capability and high speed, the resolution and contrast degrade as the imaging plane is moved deeper into the sample due to refractive index variations. We have developed a system that combines a structured illumination light sheet microscope with adaptive optics in the emission path to correct optical aberrations and increase the resolution when imaging deep into the sample. We show that our system can record neural activity fast enough to capture seizure events, and is able to correct optical aberrations throughout the sample.

We present the concept of adaptive polarization optics and validate it using a system that contains full polarization control and measurement at pixel level. Polarization control is implemented based on double pass spatial light modulator (SLM) system while the measurement is realized via an imaging Stokes polarimeter. Our design is able to extract the variation in polarization state across a beam and from this extracted information about a birefringent sample that introduces a polarization disturbance into the system. This information is used to control the SLM-based polarization state generator in order to restore the light to the desired state after passing through the sample. The method is demonstrated using a vortex half-wave retarder as the birefringent sample.

An adaptive scanning optical microscope with extended depth of field (ASOM-EDOF) is described. The system is based on the ASOM developed previously*, and uses a custom-built Thorlabs 0.21NA objective with a 75mm pupil diameter that allows scanned imaging over a circular region of 40mm diameter using fast steering mirror. The microscope is configured for fluorescence imaging with epi-illumination. A 140 actuator, 5.5µm stroke DM is conjugated to the pupil of the objective, and is used in conjunction with a Shack Hartmann wavefront sensor in an adaptive optics loop to measure and compensate errors of the objective as a function of nominal scan angle. At a given scan angle, the microscope camera forms an image of a 200µm x 200µm region with resolution of about 1.4µm. Images recorded at different scan angles can be stitched together to form a larger image mosaic. At each scan angle, the DM has been calibrated not only to compensate astatic aberrations, but also to perform an axial focal sweep: changing shape from concave to convex at high speed during a single camera exposure. This type of extended depth of field (EDOF) imaging (without aberration compensation) has been reported previously**. By combining these two techniques (ASOM and EDOF), a single recorded camera frame includes in-focus light from objects at depths from the nominal objective focus to depths +/-250µm from that focus, corresponding to an extension of the depth of focus by a factor of 100x for this microscope. The image also includes out-of-focus light from all depths. After a simple deconvolution, one can recover the in-focus light from all swept layers, condensed into a 2D image. Calibration details and performance metrics are described, along with example images from large volumetric samples. *Potsaid B, Wen JTY, “Adaptive scanning optical microscope: large field of view and high-resolution imaging using a MEMS deformable mirror,” Journal of Micro-Nanolithography Mems and Moems, [7], 10, (2008). **Shain WJ, Vickers NA, Goldberg BB, Bifano T, Mertz J, “Extended depth-of-field microscopy with a high-speed deformable mirror,” Optics Letters, [42], 995-998, (2017).

Two-dimensional spatial wavefront modulation in real-time is an essential tool for applications such as adaptive optics and laser beam shaping. Micro-mirror-based MEMS wavefront modulators have led to a major reduction in the cost of practical wavefront modulation, but the system complexities due to their reflective operation are still prohibitive. To address this issue, we demonstrate here a highly-miniaturized electrostatically actuated optofluidic transmissive phase modulator capable of positive or negative phase shifting through the use of hydromechanical coupling. The approach is based on a unique push-pull electrostatic actuation concept that exploits the inherent liquid-mechanical coupling in the design and is free of polarization and diffraction effects. This optofluidic phase modulator is able to correct aberrations up to 5th radial Zernike polynomial modes with high fidelity and, by use of sensorless wavefront estimation algorithms, allows for the realization of a completely in-line adaptive optics system.

Adaptive lenses enable compact, fast and quasi-motionless scanning in optical microscopy [1]. One drawback, however, is that the elements of an imaging system are usually optimized for a certain design-focal-length of the adaptive lens. In particular, spherical aberrations negatively influence the axial and lateral resolution as well as the signal strength in a confocal microscope. We address this problem using a novel fluid-membrane lens that is based on a piezo-glass composite membrane, where an ultrathin glass membrane is sandwiched between two piezo rings. With their two degrees of freedom, they can bend and buckle the membrane, enabling different rotated conic-section-like surfaces. An iterative control algorithm enables the simultaneous, independent tuning of the focal length and the induced spherical aberrations. We apply our adaptive lens in a confocal microscope that is extended with an additional phase measurement system to enable a wavefront-based control of the adaptive lens. Applying the aberration correction to a confocal measurement of a phantom yields an enhancement of the axial resolution improvement compared to the uncorrected measurement. To investigate the usability of the system for biological specimen, we show confocal measurements at zebrafish embryos with reporter gene-driven fluorescence in the thyroid gland. [1] Katrin Philipp, André Smolarski, Nektarios Koukourakis, Andreas Fischer, Moritz Stürmer, Ulrike Wallrabe, and Jürgen W Czarske, Volumetric HiLo microscopy employing an electrically tunable lens, Optics Express Vol. 24, Issue 13, pp. 15029-15041 (2016)

Multicore fiber bundles are attractive candidates for lensless endscopes because of their ruggedness and the relative simplicity involved in their calibration and operation. Nevertheless, the measurement of the transmission matrix of the MCF is still an interferometric process, typically requiring high stability and possibly sequential measurements. A key challenge is to replace this with a much simpler and robust technique – phase retrieval. In this talk, we will examine the major challenge in using phase retrieval in conventional MCFs. This is related to the discrete and periodic nature of the auto-correlation of an ordered MCF, resulting in the stagnation of phase retrieval algorithms in one of a multitude of local minimums. We employ phase diversity, i.e. more the one complex illumination pattern with a known phase profile can help overcome this issue. In particular, we identify that spiral phase patterns are well-suited due to their generation of complementary speckle patterns, resulting in highly non-redundant information. We experimentally demonstrate that three intensity images are sufficient to retrieve the transmission matrix with very high accuracy and success rates. Furthermore, we will also present a novel disordered MCF, which facilitates phase retrieval with a single intensity image and a priori knowledge of the core positions. This is a simple and rapidly converging method which relies on the aperiodic arrangement and the sparsity of the cores. Both these computationally inexpensive techniques highlight the potential of phase retrieval as a tool for robust phase calibration of fiber bundles in lensless imaging.

A class of miniaturized imaged systems based on multicore fibers (MCFs) and wavefront shaping, also known as lensless endoscopes, have emerged as promising candidates for non-invasive imaging deep inside the tissue. At the current stage of their development, these systems already provide features like pixelation-free wide-field and point-scanning imaging and are compatible with the majority of nonlinear imaging techniques, offering diffraction-limited resolution over a field of view only limited by the single fiber numerical aperture. Widely spaced single-mode cores in such an MCF are designed for very weak inter-core coupling, which ensures an infinite memory effect and provides good ultrashort pulse delivery framework for nonlinear imaging. In a true endoscopic setting, however, where the fiber geometry is subject to continuous deformation (at a rate from several Hz to several tens of Hz), results in inter-core phase and group delay dispersion (GDD) that changes as the fiber bends. This consequently degrades the PSF in terms of size and power in the focal spot, eventually rendering impossible to produce non-linear imaging. We addressed this issue by implementing an active measurement and compensation loop at the proximal end of an MCF, allowing to correct in real time the GDD changes and to minimize the temporal dispersion of the delivered laser pulses. We evaluate this approach against a passive scheme where a static pre-compensation is used in a compliment with a specifically designed MCF, allowing to drastically increase the MCF bending robustness.

When using multimode fibers as high-resolution endoscopes, advanced adaptive optics is needed to overcome the modal dispersion which scrambles the image. Additionally, for non-linear imaging methods, all the wavelengths of a femtosecond laser pulse must be simultaneously focused at the sample plane, with appropriate dispersion compensation, that might vary across the sample area. We investigate the bandwidth of the focused spot for a graded index fiber used as a point scanning imaging device. We demonstrate that with proper compensation for the dispersion of the spatial light modulator this can be <45 nm. We also measure the spectral phase at the sample plane, and demonstrate that this does not vary substantially with spot position.

The recent progress of controlling light propagation in complex media has enabled the use of plain multimode fiber (MMF) as compact optical endoscope with sub-micron spatial resolution and sub-millimeter footprint. With the knowledge of the MMF’s transmission matrix (TM), the scrambling of the light propagating through the fiber can be compensated, physically or computationally. Current MMF endoscopes require distal access to calibrate the TM, which furthermore is vulnerable to perturbations and bending of the MMF. This necessitates repeated TM calibration or a rigid geometry that limits the intrinsic advantages of MMF. For practical applications of MMF endoscopy, calibration of the TM should be conducted without direct access to the distal fiber end. Here, we experimentally demonstrate that the forward and backward transmission through the MMF are the transpose of each other, imposed by the laws of optical reciprocity. This results in a transpose-symmetric double-pass TM (TM2x). Although it can be readily measured from the proximal side, the symmetry prevents unambiguous deduction of the single-pass TM from the measurement of TM2x. We then propose a strategy to obtain the single-pass TM in arbitrary fiber geometry by measuring TM2x with distinct pre-calibrated distal fiber elements and discuss the necessary conditions for the pre-calibrated elements to allow recovery of the single-pass TM. This proximal calibration technique may offer a pathway to flexible MMF endoscopy and find use in related applications involving measurement of TMs.

We have developed an all-solid, step-index multimode fibre based on compound "soft-glasses" yielding a very-high NA reaching 0.96 at 1064nm. By further extending the methods of holographic control of light propagation in multimode fibres, we were able to mitigate the adverse effect of mode-dependent loss affecting the new fibre type. This enabled harnessing the full available NA almost completely, and demonstrating high-resolution focussing with output NAs up to 0.91 through lensless fibres. Further, we show that the NA and pureness of such foci allow stable three-dimensional optical confinement of micrometre-sized dielectric objects. Being inherently holographic, this technique is capable of generating an arbitrary number of optical tweezers, as well as precisely repositioning them independently in all directions. The versatility of the new instrument is demonstrated by simultaneous and dynamic 3D manipulation of large assemblies of dielectric microparticles, as well as manipulation of micro-objects inside optically inaccessible environments such a turbid cavity through an opening as small as 0.1mm. Moreover, the possibility of generating aberration-free foci with NA approaching 0.9 across the fibre core opens new perspectives for high-resolution holographic micro-endoscopy, paving the way for the delivery of advanced microscopy techniques through hair-thin fibre-optic probes.

We propose a data-driven approach for light transmission control inside multimode fibers (MMFs). Specifically, we show that a convolutional neural network is able to reconstruct amplitude/phase modulated images from scrambled amplitude-only images obtained at the output of a 0.75m long MMF with a fidelity (correlation) as high as ~98%. We show that the trained network shows good generalization as well. In particular, it is shown that the network is able to reconstruct images that do not belong to train/test datasets.

Focusing light through scattering materials by modifying the phase profile of the illuminating beam has attracted a great deal of attention in the past decade paving the way towards novel applications. Here we introduce a tradeoff between two seemingly independent quantities of critical importance in the focusing process: the size of the focal point obtained behind a scattering medium and the maximum achievable brightness of such focal point. We theoretically derive and experimentally demonstrate the fundamental limits of intensity enhancement of the focal point and relate them to the intrinsic properties of the scattering phenomenon. We demonstrate that the intensity enhancement limitation becomes dominant when the focusing plane gets closer to the scattering layer thus limiting the ability to obtain tight focusing at high contrast, which has direct relevance for the many applications exploring scattering materials as a platform for high resolution focusing and imaging

After travelling through a disordered medium an ultrashort pulse of light gets completely scrambled both temporally and spatially. Multiple scattering tends to elongate the pulse duration and to distort it spatially, as each spectral component of the pulse generates a different speckle pattern. This mixing results in a complex spatio-temporal speckle pattern. By determining the Multi Spectral Transmission Matrix (MSTM) of the medium, one can achieve full control of transmitted light both in time and space only by exploiting spatial degrees of freedom of a single SLM. This operator is a stack of monochromatic Transmission Matrices (TM), measured for all the spectral components of the pulse. Although this technique has proved its efficiency, its technical complexity precludes its broad dissemination. Primarily, it requires a pulsed laser capable of tunable CW operation to measure all monochromatic TMs. Additionally, the scattering medium needs to be stable during the whole experiment, which is a major limitation for thick media with large number of independent TMs. Here, we report a new technique to parallelize the full MSTM measurement of a highly scattering medium. It speeds up acquisition an order of magnitude, and does not require a tunable source. To this end, a micro-lens array and a diffraction grating are used to encode both spatial and spectral information of the output speckle on a single CMOS camera. We experimentally demonstrate the advantage of the technique by measuring MTSM in very strong scattering regime (N_λ>30) where the conventional method would be impractical.

The manipulation of optical beams through a scattering medium using the measurement of its transmission matrix (TM) offers an exciting range of opportunities, from fundamental understanding of disordered media to information processing and optical imaging. The use of the TM to create a focus through a scattering medium is a key element to produce nonlinear signals for in-depth imaging in biological media. Measuring a TM by wavefront shaping and self-reference interferometry in a broadband spectral regime (typically from a 100 fs pulsed laser) permits moreover to coherently gate the incident beam so that the pulse stays inherently short enough to produce nonlinear signals. The TM knowledge moreover permits to access point scanning modality for imaging large field of views. It is still however very challenging to perform nonlinear imaging by frequency mixing through a scattering medium using this method: first because scanning a refocus exhibits non homogeneous features of the reference speckle, and second because the TM is valid only for a limited range of wavelengths which is well below the wavelength differences used in sum frequency generation, four wave mixing or Coherent Anti Stokes Raman Scattering (CARS). In this work, we exploit the angular and spectral correlation properties of the TM to produce fast sum frequency mixing imaging of large field of views through up to millimeter thick highly scattering biological tissues, such as mice spinal cords. This work opens new prospective for chemically specific imaging modalities in tissues such as CARS.

Optical wavefront shaping is a powerful technique to control the distribution of light in the focus of a microscope. Combined with optogenetics, it holds great promise for a precise manipulation of neuronal activity with light. A better understanding of complex brain circuits however, requires advanced and flexible optical methods capable of simultaneously photo-exciting multiple neurons, possibly using dedicated excitation shapes, arbitrarily distributed in the three dimensions, with single-cell resolution. At the same time, the study of deep brain structures with all optical techniques, even in the multi-photon regime, is limited by scattering to a depth of few hundreds µm. Here we first present a new optical scheme, based on the spatio-temporal shaping of a pulsed laser beam, to project several tens of spatially confined two photon excitation patterns in a large volume. Using two spatial light modulators and the temporal focusing technique we are able to produce at least 4 different extended excitation patterns, with single cell axial confinement, that we independently multiplex at the sample volume an arbitrary number of times. We fully characterise the optical response of the system, discuss the possibility of simplifying it at the expenses of flexibility, and subsequently exploit it to perform multi-cell volumetric excitation in both Drosophila and zebrafish larvae. Finally, we summarise our recent efforts towards the extension of such method to a micro endoscope, which could be used for the study of complex neural circuits in deep brain structures, thus overcoming the limitations imposed by scattering.

Scattering of light limits the depth at which a focus can be formed in turbid media. However, light can be focused through and inside thick and strongly scattering samples by spatially modulating the incident light using wavefront shaping techniques [1]. Wavefront shaping requires feedback from a localized reporter, for example, a point detector or a fluorescent ‘guide star’ inside the sample. In some cases, a localized reporter is not available as a source of feedback. For instance in multiphoton fluorescence microscopy, the only available feedback signal is the total fluorescent signal coming from inside the sample. Even with this non-localized form of feedback, Katz et al. [2] were capable of forming a single diffraction-limited focus behind a strongly scattering layer. However, the statistics behind this nonlinear optimization procedure are poorly understood, and the location at which this blind focusing method will form a focus could not be predicted or controlled. We developed an analytical model to predict the outcome of the blind focusing method. Our model allows us to determine under which conditions the optimization algorithm converges to a single diffraction-limited focus, and how the location of this optimized focus can be controlled. Furthermore, we can find the parameters that determine the convergence rate of this blind focusing procedure. The model is validated with experiments through strongly scattering samples, and an excellent agreement was found. [1] I.M. Vellekoop, Optics Express 23, 1-18 (2015) [2] O. Katz, E. Small, Y. Guan, Y. Silberberg, Optica 1, 170-174 (2014)

In contrast to conventional imaging systems that map an object point by point, measurements with random sensing functions in combination with computational reconstruction may afford novel imaging architectures. Here we demonstrate imaging of axial reflectivity profiles using random temporal-spatial encoding created by modal interference in a multimode fiber (MMF). Light from a broadband source (∆λ = 60nm) centered at 1310nm is split into a sample and a reference arm. In the sample arm, light in a single spatial mode is reflected by the axial reflectivity profile of the sample and coupled back into the same spatial mode. The reference light propagates through a MMF and interferes with the sample light in an off-axis geometry on a camera for holographic recording. Since the MMF supports various guided modes with distinct propagation constants, the short-coherence sample light only interferes with the spatial modes of the reference light that have matching path length. During an initial calibration procedure, interference patterns of a mirror reflection in the sample arm are recorded for varying axial mirror positions. Once this random sensing matrix (RSM) is established, the axial reflectivity profile of an object in the sample arm can be reconstructed from a single interference pattern by the multiplication with the inverse of RSM. By using a 2m long 0.22 NA MMF and tailoring the coupling regime within the MMF, we achieved axial ranging more than a centimeter. Flexible integration of polarization sensing or multi-focus imaging in a single snapshot could be envisioned in this random imaging architecture.

Propagation of coherent light through a disordered medium generates a speckle pattern at the output, due to light scrambling by multiple scattering events. Over the last decade, wavefront shaping techniques have opened a new way to perform imaging through disordered systems, using spatial light modulators (SLM). In particular, the optical transmission matrix (TM) links the input field, modulated by the SLM, to the output field. It enables arbitrary diffraction limited spatial focusing of light after propagation in the medium. We recently extended this technique to arbitrary focus shape, by numerically computing a mask (amplitude and/or phase) onto a virtual pupil accessible via a Fourier transform operation. This TM method has also recently been extended to the spectral domain using the Multi-Spectral Transmission Matrix (MSTM). MSTM is a stack of monochromatic TMs for different wavelengths of a broadband pulse. In conjunction with a SLM, this operator makes a scattering medium behave like a arbitrary dispersive element or pulse shaper. Here, we generalize our PSF-control ability to the spectral domain, we experimentally demonstrate focusing a broadband pulse with a different PSF for two different wavelengths domains. Controlling these spectrally dependent k-spaces opens also interesting perspectives to achieve Temporal Focusing (improved depth sectioning for multiphoton excitation), at the output of thick scattering medium. We present numerical simulations of such temporal focusing, and study the advantages of our technique compared with conventional temporal focusing methods.

To control the light propagation in turbid media, it is necessary to reconstruct the output wavefront. In 2007 Vellekoop et al.¹, developed an iterative algorithm that divides the input wavefront in N x N channels called segments, after passing through turbid media, the output wavefront is reconstructed by measuring the intensity at a desired point, and then the phase of each channel is updated, the final N x N phase is called optimal phase matrix. The interpolation technique is capable of transforming a N x N matrix into a second 2N x 2N matrix, where, the 50 percent of the resulting matrix elements correspond to the homogeneous distribution of the original matrix values and the remain values are generated by interpolating the neighbors. Our proposal uses the optimal phase matrix obtained by an iterative algorithm, and then the number of segments is increased by interpolation. We analyze the circularity, the signal to noise ratio (SNR), the Full Width at Half Maximum (FWHM) and the correlation for different output wavefronts obtained by the optimal phase matrix and the interpolation optimal phase matrix. Our results show that, Circularity, SNR, and FWHM parameters do not change significantly and the acquisition time of the optimal phase matrix decreases compared with a similar matrix obtained by the iterative algorithm; therefore, our proposed technique that consists in the combination of interpolation and iterative algorithm is useful to study the light transmission in turbid media when a high resolution is needed in the transmission matrix, for example, phase holograms transmission through turbid media.

Light scattering in complex media is a pervasive problem across many areas, such as deep tissue imaging, and imaging in degraded environment. Major progress has been made by using the transmission matrix (TM) framework that characterizes the “one-for-one” input-output relation of a fixed scattering medium as a linear shift-variant matrix. A major limitation of these existing approaches is their high susceptibility to model errors. The phase-sensitive TM is inherently intolerant to speckle decorrelations. Our goal here is to develop a highly scalable imaging through scattering framework by overcoming the existing limitations in susceptibility to speckle decorrelation and SBP. The proposed model is built on a deep learning (DL) framework. To satisfy the desired statistical properties, we do not train a convolutional neural network (CNN) to learn the TM of a single scattering medium. Instead, we build a CNN to learn a “one-for-all” mapping by training on several scattering media with different microstructures while having the same macroscopic parameter. Specifically, we show that our CNN model trained on a few diffusers can sufficiently support the statistical information of all diffusers having the same mean characteristics (e.g. “grits”). We then experimentally demonstrate that the CNN is able to “invert” speckles captured from entirely different diffusers to make high-quality object predictions. Our method significantly improves the system’s information throughput and adaptability as compared to existing approaches, by improving both the SBP and the robustness to speckle decorrelations.

Wavefront sensing is typically accomplished with a Shack-Hartmann wavefront sensor (SHWS), where a CCD or CMOS is placed at the focal plane of a periodic, microfabricated lenslet array. Tracking the displacement of the resulting spots in the presence of an aberrated wavefront yields measurement of the relative wavefront introduced. A SHWS has a fundamental tradeoff between sensitivity and range, determined by the pitch and focal length of its lenslet array, such that the number of resolvable tilts is a constant. Recently, diffuser wavefront sensing (DWS) has been demonstrated by measuring the lateral shift of a coherent speckle pattern using the concept of the diffuser memory effect. Here we demonstrate that tracking distortions of the non-periodic caustic pattern produced by a holographic diffuser allows accurate autorefraction of a model eye with a number of resolvable tilts that extends beyond the fundamental limit of a SHWS. Using a multi-level Demon’s image registration algorithm, we are able to demonstrate that a DWS demonstrates a 2.5x increase in number of resolvable prescriptions as compared to a conventional SHWS while maintaining acceptable accuracy and repeatability for eyeglass prescriptions. We evaluate the performance of a DWS and SHWS in parallel with a coherent laser diode without (LD) and with a laser speckle reducer (LD+LSR), and an incoherent light-emitting diode (LED), demonstrating caustic-tracking is compatible with coherent and incoherent sources. Additionally, the DWS diffuser costs 40x less than a SHWS lenslet array, enabling affordable large-dynamic range autorefraction without moving parts.

In computational imaging by digital holography, lateral resolution of retinal images is limited to about 20 microns by the aberrations of the eye. To overcome this limitation, the aberrations have to be canceled. Digital aberration compensation can be performed by post-processing of full-field digital holograms. Aberration compensation was demonstrated from wavefront measurement by reconstruction of digital holograms in subapertures [Kumar, A. et al. Optics express 21.9 (2013): 10850-10866.], and by measurement of a guide star hologram [Liu, C. et al. Applied optics 52.12 (2013): 2940-2949.]. Yet, these wavefront measurement methods have limited accuracy in practice. For holographic tomography of the human retina, image reconstruction was demonstrated by iterative digital aberration compensation, by minimization of the local entropy of speckle-averaged tomographic volumes [Hillmann, D. et al. Scientific reports 6 (2016): 35209.]. However image-based aberration compensation is time-consuming, preventing real-time image rendering. We are investigating a new digital aberration compensation scheme with a deep neural network to circumvent the limitations of these aberration correction methods. To train the network, 28.000 anonymized images of eye fundus from patients of the 15-20 hospital in Paris have been collected, and synthetic interferograms have been reconstructed digitally by simulating the propagation of eye fundus images recorded with standard cameras. With a U-Net architecture [Ronneberger, O. et al. International Conference on Medical image computing and computer-assisted intervention. Springer, Cham (2015): 234-241.], we demonstrate defocus correction of these complex-valued synthetic interferograms. Other aberration orders will be corrected with the same method, to improve lateral resolution up to the diffraction limit in digital holographic imaging of the retina.

Multimode fiber imaging using transmission matrix (TM) is a promising way of deep inspection of living objects. These techniques, however, face a problem of obtaining the transmission matrix using phase-shifting interferometry with external reference beam, which requires additional instrumentation and increases space requirements of the experiment. We suggest a method employing input mode represented by a focal spot at the proximal end of the fiber as an internal reference wave. Due to speckle nature of the output, it is necessary to cover blind spots in the transmission matrix arising from the lack of interference signal by several measurements with different input modes used as a reference. The effect of optimized selection of internal references will be analyzed and compared with the external reference approach.

The reconstruction of objects hidden behind scattering media has been partly triggered by the striving for applying those techniques in biological imaging where fluorescence serves as an image contrast. Object reconstructions rely on speckle correlations that are rooted in the spatial and spectral memory effect of the medium. In the majority of previous seminal experiments in the field, laser light transmission was used as object contrast and millimetric sized objects were separated by cm distances from the scattering media. An open question therefore was to what extent microscopic-sized fluorescent structures can be reconstructed at the proximity and behind a scattering media, in a conventional epi wide field microscope as found in imaging laboratories. Here, we make use of the magnification properties of the scattering medium itself to resolve a scattered fluorescent image on a sCMOS camera. Due to the epi configuration, light is scattered in the excitation as well as the detection. Hence, we show that the use of the full fluorescence bandwidth greatly facilitates the detection of scattered signals in a single-shot at short integration times. Along those lines we present a phase retrieval algorithm of Fienup-type that is less sensitive to low signal conditions and takes four times less iterations than comparable phase retrieval algorithms. Backed by a complementary Fourier smoothing image processing technique that no longer requires an autocorrelation of the scattered light image to separate out the speckle noise, we are able to provide high quality micrometric fluorescent object reconstructions in a simple conventional inverted microscope.

A revised version of

Yunqi Luo, Huanhao Li, Ruochong Zhang, Puxiang Lai, and Yuanjin Zheng "Deep learning assisted optical wavefront shaping in disordered medium", Proc. SPIE 10886, Adaptive Optics and Wavefront Control for Biological Systems V, 1088612 (20 February 2019); doi: 10.1117/12.2504425 was published on 17 April 2019.

Details of the revision are provided in the text that accompanies this Erratum. The original paper has been updated.

Strong scattering in turbid medium is a severe difficulty for optical imaging and measurement in an animal body. Therefore, the suppression of scattering effect is crucially important in many applications of biomedical optics. In the optical transillumination imaging of an animal body, this effect appears as strong blurring of the image. This blurring poses a fundamentally important difficulty restricting the practical application of transillumination imaging. Therefore, we have attempted to suppress scattering effect using the time-reversal ability of phase-conjugate light. For our previous study, we constructed a digital system to generate light not only with conjugated phase but also with the same intensity distribution as non-scattered signal light. Using this system, we were able to restore the pattern of incident light from the blurred image because of time-reversal propagation of the phase-conjugate light. In comparison to a case with phase information only, we found that addition of the intensity information greatly improves the scattering suppression capability of the time-reversal principle. However, our pilot study showed this ability was valid only for the scattering medium with the optical distance OD less than 1. This report describes the improvement of our measurement system to make the scattering suppression possible for turbid media with OD of more than 1 using a light source with a longer coherence length. Scattering suppression was effective for spatial frequencies of 0.4–1.0 lp/mm through the scattering medium up to OD=1.8.

Adaptive optics (AO) is a promising technique for correcting wavefront errors induced by complex structures of biological samples which significantly causes image degradation. We develop a microscopic AO system with a Shack-Hartmann wavefront sensor based on image correlation. The correlation-based wavefront sensing is feasible using an extended object under bright-field illumination, as well as spot fluorescence. To make the correlation-based sensing more reliable, we newly introduce a technique of excluding sub-images with insufficient quality. We show experimental results under a variety of conditions for objects, light sources, and wavefront error sources. In any cases, we confirmed that the AO system effectively worked so as to improve image qualities.

The ability to focus laser beam with wavelength 0.65 um through the multiply scattering suspension of polystyrene microspheres, diluted in distilled water, was investigated. Experimental setup, contained the Shack-Hartmann sensor for measurements of the local slopes of the Poynting vector, the CCD camera for estimation of the far-field focal spot’s intensity and the bimorph mirror with 48 electrodes was built. Numerical and experimental investigations of focusing efficiency was carried out also.

We introduce an optimization-based open-loop control method for 2D wavefront modulators. The optimization problem is convex with inequality constraints and can be solved using an interior-point method in real-time. Compared to conventional influence matrix inversion, this new method takes into account the system limitations, such as the actuation polarity and voltage limits of the drivers. It searches for the global optimum of actuation signals within system boundary constraints. Consequently, while reducing the complexity of the hardware, it is more immune to systematic errors and guarantees optimality of the actuation signals. The control system is implemented on two different electrostatically-actuated phase modulators; a conventional deformable mirror and a novel refractive optofluidic phase modulator. We experimentally compare the performance of the optimizationbased controller with conventional methods for high order Zernike mode replication. It is demonstrated that the introduced technique enables more accurate control for both modulators, particularly at large correction amplitude and/or higher order corrections.

Inner diffraction phenomenon is known as the major obstacle of light transmission through scattering media such as ground glasses, skin or biological tissue. Recently, the most effective and convenient solution is wave-front shaping technique which modulates the field profile of incident light by using a spatial light modulator (SLM). For practical and advanced biomedical applications, requirement of speedy response, high accuracy and large energy delivery are necessary. In our previous work, we presented a wave-front shaping technique and utilized optical memory effect for swiftly drawing various 2 dimensional (2D) shapes or contours through a scattering medium without any mechanical movement. However, with process of scanning angle phase profile and shifting phase pattern on SLM, the accuracy and beam energy utilization are still very much restricted. Here, we present an exceeding improvement from previous technique by establishing optical conjugate planes between SLM and scattering medium surface, which is also known as 4F system. With only one phase profile for creating a focus spot behind scattering medium, we are able to swiftly move focus spot in 3D space or draw any 3D contours through turbid medium without scanning or shifting process. The new approach allows us to deliver laser energy through a scattering medium to any spot within 3D memory effect space with very fast response, high accuracy and more importantly, fully utilized laser beam energy. Our approach demonstrates a practical method to control light through scattering media for prominent applications such as opto-genetic excitation, minimal invasive laser surgery and other related fields.

A revised version of this paper was published on 17 April 2019. Details of the revision are provided in the text that accompanies this Erratum. The original paper has been updated.