1. Introduction

Endoscopes rely on minimally-invasive elements to deliver and collect signals from areas of difficult access in a specimen. Imaging of inner parts of bulk organs demands little-invasive probes that allow investigation of tissue at adequate penetration depths [1–4]. Particularly when it comes to the brain, the study of neuronal activity in animals behaving, operating, and moving in a natural fashion is of upmost importance. The miniaturization of the available imaging tools gave room to the first observations of complex neuronal networks and brain processes in vivo [5–7] and progresses in optogenetics [8]. Earliest advances utilized bundles of single-mode fibres [9], and efforts are now made in the direction of developing a probe with the smallest footprint so far: a novel class of ultra-narrow endoscopes comprised of a single multimode fibre (MMF). By regarding multimode fibres as analogous to optically turbid media [10,11], light transport can be controlled using spatial light modulators and by employing recently formulated holographic methods [12–16].

In spite of the reduced footprint of a single fibre, the insertion of the probe in tissue modifies it upon insertion by dragging and straining the medium around it, causing mechanical tear of cells, and potentially inflammation [17]. The imaging fibre facet affects particularly the cells under study, which can lead to ambiguities. The issue can be circumvented by diminishing the probe’s overall diameter, which comes at cost of field of view (FOV) size, or by post-processing the fibre termination, for instance, into a flat-cone [6,18]. The latter alternative allows smoothing the penetration inside tissue while keeping the core, and therefore the FOV, intact. But the distal fibre facet still contacts directly with the tissue under investigation.

Following the direction of higher-fidelity imaging, we propose a fibre probe that constitutes, to the best of our knowledge, the first all-fibre side-view microendoscope. The design comprises a ${110}\;\mathrm{\mu} \textrm {m}$-diameter multimode fibre, whose distal end was mechanically modified, conveying to the probe capabilities of off-optical-axis imaging [19–21] [Fig. 1(a)]. By performing imaging perpendicularly to the direction of fibre insertion, we anticipate cells to remain unperturbed across the FOV. As a mere guide to the eye, we replicated the perturbation of cells in a sample upon insertion of an optical fibre in a conventional straight-view configuration, and in a side-view configuration [Figs. 1(d) and 1(e) and Visualization 1). The model considers cells with randomly distributed volumes assigned a priori, acting forces only upon their neighbors. The forces have magnitudes proportional to the change of the cell’s volume, and direction perpendicular to the cell walls. No tangential forces are considered, therefore cells can slip freely within the tissue structure. In each iteration the probe is inserted deeper into the tissue, displacing cells in front of the facet, thereby causing a propagating perturbation. The colour of the cell walls corresponds to the accumulated displacement of the cell’s position with respect to its initial neighbors.

 

Fig. 1. (a) Schematic illustration of the conventional ‘straight-view’ and novel ‘side-view’ single-fibre imaging probes, with indication of the angles of the polished surfaces and focal planes (FPs). In a straight-view configuration, light propagates along the fibre core and is focused on the focal plane, whose distance to the output facet is set during a calibration procedure. For side-view imaging, light is internally reflected on a ${45}^{\circ }$-polished surface with a reflective coating, leaving the fibre through the flat ${5}^{\circ }$-polished output facet. (b) and (c) Images of a side-view probe: lateral view (b) and ${5}^{\circ }$ output facet (c). (d) and (e) Illustration of damage caused on tissue upon insertion of a straight-view probe (d) and a side-view probe (e) inside an elastic medium divided into cell units. The colour bar indicates the displacement of each cell unit with respect to its initial position prior the insertion of the fibre. In the latter case (e) the cell units suffer little perturbation around the probe’s surface, particularly the ones directly in the field of view (see Visualization 1).

Download Full Size | PDF

Through a simple developed ray optics model, the novel probe was numerically characterized in terms of its numerical aperture (NA) dependency on specific foci positions in the FOV. We compared focusing performance of both imaging modalities regarding power in the focus and numerical aperture, as well as imaging performance using a 1951 USAF resolution test chart. Lastly, we created a tissue phantom featuring fluorescent beads to study the motion of particles upon fibre insertion.

2. Methods

2.1 Preparation of fibre probes

Both straight- and side-view probes were prepared using the same commercially available step-index MMF (Ceramoptec Optran WFGe) with core and cladding diameters of ${100}\;\mathrm{\mu} \textrm {m}$ and ${110}\;\mathrm{\mu} \textrm {m}$, respectively, and 0.37 NA. At the working wavelength of ${488}\;\textrm {nm}$ used in this study, this fibre supports approximately 28 000 waveguide modes, i.e., 14 000 per orthogonal polarization state. As a first step, the fibre is stripped from its acrylate coating, followed by a flat cleave to both its extremities. For the straight-view probe, the fibre is not processed any further.

For the side-view probe, the distal fibre facet is subsequently mechanically polished at ${45}^{\circ }$, rotated by ${180}^{\circ }$ around its longitudinal axis and polished at ${5}^{\circ }$, removing part of the cladding and about ${17}\;\mathrm{\mu} \textrm {m}$ in the core [Figs. 1(b) and (c). These parameters were selected taking into account the trade-off between two considerations. First, the FOV should provide a suitable size comparable to that of the straight-view probe; second, when creating the shallow ${5}^{\circ }$-slant, core material should be polished off as little as possible to preserve the propagating modes. A custom-assembled polishing system, that allows for real-time inspection of the polishing procedure, was built for the purpose of developing this specific type of probe. Polishing is achieved by bringing the fibre tip to contact with a polishing film attached to a fast-spinning disk. The polishing film has a layer of diamond abrasive particles, and the choice of grain size depends on the stage of polishing. Starting off with a large grain size (${6}\;\mathrm{\mu} \textrm {m}$-diameter particles), a rough and fast polishing is achieved, after which a ${1}\;\mathrm{\mu} \textrm {m}$ grain size is used for better surface quality, and finally a polishing paper with ${0.02}\;\mathrm{\mu} \textrm {m}$-diameter silicon dioxide particles was used for the finest polishing. The ${45}^{\circ }$ surface is then coated with a reflective aluminium layer using a vacuum evaporator (JEOL JEE-420).

Both side- and straight-view probes used in the following experiments are approximately ${25}\;\textrm {mm}$ long.

2.2 Imaging system

To assess and compare the focusing and imaging performance of the novel side-view probe and the standard straight-view probe, we assembled an imaging system based on Ref. [6], depicted schematically in Fig. 2. This system allows first characterizing the light propagation through the multimode fibre probes by measuring the optical transmission matrix [13,22,23]. This is performed in the calibration units presented in the scheme.

 

Fig. 2. Optical setup for both straight- and side-view imaging probes. The calibration units are used to acquire the transmission matrix of each probe prior imaging. Following the calibration, a series of holograms is applied to a digital micromirror device, shaping the incoming laser beam. The modified wavefronts are coupled into the proximal fibre facet, which translates into focal points scanned across the distal end of the probe. Through the same fibre facet, the fluorescence signals emitted by a distal object are collected and detected by a photomultiplier tube on the proximal side, allowing reconstruction of an image. Legend: MMF, multimode fibre; DMD, digital micromirror device; ID, iris diaphragm; L, lens; HWP, half-wave plate; QWP, quarter-wave plate; BS, non-polarizing beamsplitter plate; PBS, polarizing beamsplitter cube; MO, microscope objective; Cam, camera; S, shutter; OI, optical isolator; M, mirror; DM, dichroic mirror; F, filter; PMT, photomultiplier tube detector. The detailed list of components can be found in the Appendix.

Download Full Size | PDF

A collimated laser beam with ${488}\;\textrm {nm}$ wavelength illuminates the surface of a digital micromirror device (DMD) under an incidence angle of ${\approx 24}^{\circ }$. The DMD is a binary-amplitude modulator comprised of a 2D array of micromirrors which can individually and selectively be tilted by ${\pm 12}^{\circ }$. The chosen incidence angle ensures that micromirrors in the ‘on’ state ($+{12}^{\circ }$) reflect light in the direction of the succeeding optical components aligned with the DMD normal axis, and micromirrors in the ‘off’ state (${-12}^{\circ }$) deviate light towards a beam dump – on account of simplicity, the DMD in Fig. 2 is not depicted with the correct orientation. Employing the DMD in the off-axis regime using the Lee hologram method, allows to use the DMD as a phase modulator [24,25]. Lens pair L1+L2 forms a telescope to expand the beam so to overfill the DMD. Half-wave plate HWP1 together with polarizing beamsplitter cube PBS1 control the total transmitted power to the system, while half-wave plate HWP2 and polarizing beamsplitter cube PBS2 control the fraction of power shared between the signal and reference beams. Lenses L3-L6 form two 4f systems to relay the DMD holograms to the proximal fibre facet, and demagnify them to fit the fibre core. Only the first diffraction order is coupled into the MMF. The polarization state is governed by the pair half-wave plate HWP3 and quarter-wave plate QWP1, which guarantee circular polarization of the input fields. Regardless of the spatial distribution of light, it has been shown that circular polarization is preserved very well in step-index multimode fibres [26]. The metal reflective layer of the side-view probe however causes phase shifts between the s- and the p-polarization components, which result in deviations from the circular polarization state of the outputs. Out of all available choices of metal material, which lend themselves to coating depositions, the aluminium causes the least polarization variations. As indicated by the results below, this effect does not influence the imaging performance. In the straight-view configuration, microscope objective MO1 and tube lens L7 image the MMF distal endfacet onto camera Cam1. Similarly and alternatively, in the side-view configuration the image is relayed to camera Cam2 by microscope objective MO2 in combination with tube lens L8.

The transmission matrix is acquired through phase-shifting interferometry, by interfering the signal beam emerging from the fibre probes with an external phase reference, whose polarizations are aligned by half-wave plate HWP4. The beams are combined either by non-polarizing beamsplitter plate BS2, in the side-view configuration, or by non-polarizing beamsplitter plate BS1, in the straight-view configuration. To reach the highest fraction of power in each focal point, the transmission matrix was measured using an oversampled basis of input fields [27], comprised of approximately 21 000 truncated plane-waves with varying $k$-vectors at the proximal fibre facet. As the output basis, a square grid of approximately 100 000 focal points were used, which are conjugate to camera pixels. During this calibration procedure, for each input field the interferometric response is measured at the distal calibration plane for all output modes simultaneously. Once the transmission matrix has been acquired, the sequence of binary holograms that generate a focus at each calibrated position on the distal calibration plane is calculated and uploaded to the DMD on-board memory. By exposing an object in the sample plane to this sequence of foci, an image can be reconstructed from the light signals which are collected back by the fibre. During imaging, a mechanical shutter S blocks the reference beam.

To assess the imaging resolution attainable with each fibre probe, a negative 1951 USAF test chart was placed on the calibration plane on the distal side. Images were reconstructed from the transmitted intensity reaching a photodiode placed on a plane conjugate to the test target, i.e., in place of Cam1 in the straight-view configuration, or in place of Cam2 for side-view.

When imaging fluorescent beads, the emitted signal back-propagating through the fibre (orange beam) is spectrally isolated from the excitation signal by dichroic mirror DM and long-pass filter F, and subsequently detected by photomultiplier tube detector PMT.

In all experiments, calibration (i.e., acquisition of the transmission matrix) was carried out in air, in a plane located ${20}\;\mathrm{\mu} \textrm {m}$ away from the distal facet of the probes.

3. Results

3.1 Modelling the imaging performance of modified probes

The principles of holographic endoscopy enable synthesizing any desired optical field within the available field of view and range of accessible spatial frequencies. Finding these constraints then allows estimating the local focusing ability, which projects itself to imaging resolution and contrast. The approximative model used here associates the spatial spectrum components of each focal position across the imaging plane with rays converging into the focal point under their corresponding angles. These are followed backwards, and their reflections and refractions are evaluated at each interface they encounter utilizing Fresnel equations and Snell’s law (the reflective mirror surface is considered lossless). All such resulting secondary rays are followed until they either meet conditions for fibre guiding (i.e., they reach the fibre core of the unmodified part of the probe, with propagation angle below the critical value for the core/cladding interface) or they leave the probe. Furthermore, if the power of any secondary ray becomes reduced below $10^{-4}$ of the power of the initial field, it is excluded from the following evaluation and its power is considered as a loss. Finally, the model does not follow rays in both polarization states, utilizing instead mean values of power split between the reflections and refractions obtained for the s- and p-polarization states.

Figure 3(a) shows a small selection of the $k$-space components forming a randomly chosen focal point in the image plane, which have been found to support propagation in the fibre. Examples of resulting apertures in $k$-space are presented in Fig. 3(b) for foci formed at different locations. Despite all secondary rays being followed until minuscule power fractions, the apertures appear almost binary. The modified termination therefore acts as a hard and position-dependent aperture in $k$-space. These apertures feature a central circular area with dimensions corresponding to the NA of the bare-terminated fibre and additional side lobes. These side lobes modify the point spread function (PSF) locally, enabling higher resolution, yet manifesting inhomogeneity and anisotropy across the FOV. The PSF at 2500 different positions is shown in Fig. 3(c), where yellow circles highlight those corresponding to the spatial spectra illustrated in Fig. 3(b). The imaging performance utilizing the full spatial spectrum is shown in Fig. 3(d), where groups 9 to 12 of a negative 1951 USAF resolution test chart are tiled across the ${100}\;\mathrm{\mu} \textrm {m}$-wide FOV.

 

Fig. 3. Modelling the focusing and imaging performance of the side-view imaging probe. (a) Ray tracing of $k$-vectors which meet the guiding conditions of the fibre, converging into a single focus in the focal plane. (b) Position-dependent aperture in $k$-space, corresponding to the point spread function at different selected locations; ${k_x}/{k}$ and ${k_y}/{k}$ represent transverse spatial frequencies normalized to the wavenumber in free space. (c) Average intensity projection of the PSF at different positions in the focal plane; the locations highlighted correspond to those shown in (b). (d) Imaging of a negative 1951 USAF resolution test chart tiled over the whole field of view.

Download Full Size | PDF

3.2 Focusing performance

Following the calibration step above, we sequentially generate distal focal spots across the whole FOV, and build a high-dynamic range (HDR) intensity distribution for each focus by combining 8-bit depth images with exposure times of ${59}\;\mathrm{\mu} \textrm {ms}$, ${590}\;\mathrm{\mu} \textrm {ms}$, ${5900}\;\mathrm{\mu} \textrm {ms}$, ${59000}\;\mathrm{\mu} \textrm {ms}$, and ${590000}\;\mathrm{\mu} \textrm {ms}$. An HDR darkframe image was obtained in the same way by blocking the laser beam, and the averaged background level was subtracted from each HDR intensity distribution for all foci.

By fitting each focus to an Airy disc function, we estimate the spot size as the diameter of the fitted distribution where its intensity is ${1}/{e^2}$ of its maximum value, which in general yielded smaller values for the side-view probe [Figs. 4(a) and 4(d)]. The effective NA was retrieved from the spot size of the foci, and therefore shows an overall higher NA for the side-view probe [Figs. 4(b) and 4(e)]. The estimated fibre NA, i.e., the one measured with the straight-view probe, reaches a maximum value of about 0.37, same as the nominal NA of the fibre. For the side-view probe, the effective NA reaches a value of around 0.47, exceeding the nominal NA. We attribute this difference to the conversion of waveguide modes into higher spatial frequencies along the region where the fibre core has its size reduced by the ${5}^{\circ }$-polished slant, a phenomenon also known in tapered optical fibres [28]. These findings agree with the ray-tracing model [Fig. 3(b)], where we find additional side-lobes that are not found in a straight-view probe. The dependence of the aperture’s profile on the position of the PSF translates into the apparent inhomogeneity of the NA across the side-view FOV [Fig. 4(e)]. The difference in the field of view shape for both fibre probes is immediately discernible: for straight-view, the FOV corresponds to the cross section of the fibre core, i.e., a circular region with an approximately ${100}\;\mathrm{\mu} \textrm {m}$-diameter; for side-view, the FOV is slightly asymmetric, owing to the asymmetric probe geometry at its termination, shown above in Fig. 1(c). The area of the FOV can be adjusted for a specific working distance during calibration, and hence a circular, smaller, and therefore more uniform FOV can be selected. To clearly evince the attainable FOVs of each probe, for the measurements in Fig. 4 we performed the calibration for the largest possible area in the distal plane.

 

Fig. 4. Assessment of focusing performance over the field of view (FOV) of the straight-view (a-c) and the side-view (d-f) fibre probes. (a) and (d) Spot size of the fitted Airy disc distribution to each captured focus, taken as the diameter where their intensities reach ${1}/{e^2}$ of the maximum value. (b) and (e) Effective NA retrieved from the spot sizes. (c) and (f) Fraction of optical power contained in each focus. The FOV takes the shape of the distal fibre facet, reason behind the circular-shaped FOV for the straight-view probe and an irregular shape for the side-view probe. The diameter of the scanned foci has a size of $(1.1\pm 0.2)\;\mathrm{\mu} \textrm {m}$ for side-view and $(1.3\pm 0.3)\;\mathrm{\mu} \textrm {m}$ for straight-view, and therefore the obtained effective NA value is higher for the side-view probe $(0.37\pm 0.06)$ than for the straight-view fibre $(0.32\pm 0.05)$.

Download Full Size | PDF

Additionally, we evaluated the fraction of power contained in each individual focus over the total output power emerging from the fibre on its distal termination. This power ratio, depicted in Figs. 4(c) and 4(f), quantifies how much optical power is indeed controlled and thus contributes to the generated foci, with the remaining giving rise to an uncontrolled speckled background spreading across the FOV. The measured maximum value was around 0.48 for straight-view and 0.57 for side-view, and the spatial distribution in both cases is not uniform over the entire FOV. Both probes are fabricated from the same fibre, and therefore propagate the same number of modes. Due to the smaller dimensions of the side-view FOV, the power ratio distribution is overall higher in comparison to the straight view probe.

3.3 Imaging performance

To determine the attainable imaging resolution using both straight- and side-view probes, we imaged a negative 1951 USAF resolution test chart in transmission by scanning foci over the FOV, and reconstructed the image by integrating the overall intensity of the transmitted signal [Figs. 5(a) and 5(c)–here, only groups 8 to 10 are depicted] [29,30]. The imaging contrast for selected elements of target groups 6 to 10 is plotted in Figs. 5(b) and 5(d) as a function of their spatial frequency, expressed in terms of the wavenumber in free space. For each resolvable target element, the value of contrast was averaged over several measurements, for different locations in the FOV and for both vertical and horizontal orientations. The error bars indicate standard deviation around the mean values. The contrast data was fitted to the contrast transfer function (CTF) of a circular aperture, which describes the response of the imaging system to binary black/white bar patterns (i.e., 2D square waves) of varying periodicity (spatial frequency) [31]. This model function has two free parameters–the zero-frequency contrast and cut-off frequency–which correspond to the intersection with the vertical and horizontal axes, respectively. The latter allows estimating the effective NA of the system as half the cut-off normalized spatial frequency, which was found to be $0.35\pm 0.03$ for the straight-view probe, and $0.41\pm 0.02$ for the side-view probe. These calculated effective NA values are in good agreement with those retrieved from the foci analysis, previously shown in Figs. 4(b) and 4(e). Moreover, the difference in NA also explains why higher frequency target elements (group 10) can be resolved using the side-view probe only.

Side-view imaging presents seemingly higher contrast, which is congruent with the higher power ratio achieved with this probe [Figs. 4(c) and 4(f)].

 

Fig. 5. Assessment of spatial resolution using a negative 1951 USAF test target, imaged at ${20}\;\mathrm{\mu} \textrm {m}$ from the fibre facet. (a) and (c) Cropped images of the target using the straight-view fibre (a) and the side-view fibre (c). (b) and (d) Contrast for several elements of the USAF chart (within groups 6 to 10), as a function of their normalized spatial frequency. The error bars indicate standard deviations around the mean values. The solid lines correspond to fitted contrast transfer functions (CTFs), from which we retrieve the effective numerical apertures (NAs) as half the cut-off spatial frequencies. The effective NAs estimated in this manner are $0.35\pm 0.03$ for straight-view imaging, and $0.41\pm 0.02$ for side-view imaging.

Download Full Size | PDF

3.4 Imaging inside a fluorescence tissue phantom

To test the imaging performance of the side-view probe in a relevant environment, we prepared a tissue phantom comprising ${2}\;\mathrm{\mu} \textrm {m}$-diameter Nile red fluorescent beads suspended in a 1% agarose solution, which has been found to closely mimic mechanical properties of brain tissue [32]. The 25-mm-long probes serve the dual purpose of delivering the excitation signal to the sample, as well as collecting part of the emitted fluorescence signal, which is back-propagated along the fibre and delivered to a highly-sensitive bucket detector on the proximal side of the endoscope. During the experiments, the fibre probes were inserted deeper into the imaging phantom in ${5}\;\mathrm{\mu} \textrm {m}$ steps between each full frame acquired (see Visualization 2).

For straight-view, we observe a sudden change in the beads’ contrast as the fibre is pushed inside the sample. And while the particles appear to be static, they are in fact moving radially away from the centre of the FOV, as seen on the average intensity projection in Fig. 6(b). The cropped region of interest (ROI) in Fig. 6(c) shows in greater detail the change in contrast of a single particle as the fibre penetrates deeper into the phantom. Between consecutive frames the fibre moved ${15}\;\mathrm{\mu} \textrm {m}$. Considering that the axial extent of the foci is estimated as $\approx {6}\;\mathrm{\mu} \textrm {m}$, this indicates that the fibre probe indeed deforms the medium and drags the particles forward during insertion, reason why the particle seems to remain in focus over at least ${15}\;\mathrm{\mu} \textrm {m}$ of fibre penetration. Such deformations in the medium induced by the fibre probe constitute a difficult challenge for the reconstruction of volumetric images.

 

Fig. 6. Imaging of a sample with fluorescence beads using two imaging modalities. (a) and (d) Schematic of the straight-view (a) and side-view (d) imaging probes, highlighting the trajectories of embedded particles as the fibres penetrate deeper in the sample. (b) and (e) Average intensity projection of multiple frames taken in axial steps of ${5}\;\mathrm{\mu} \textrm {m}$, showing the actual trajectories of the particles until the moment they cease to be visible. In the polar representation shown in (b), the centre was removed for better visualization. It is seen here that the insertion of the straight-view probe pushes the particles radially, away from the centre of the FOV. Using the side-view probe, the particles move linearly along the flat ${5}^{\circ }$ window as the fibre penetrates in the sample. (c) and (f) Cropped ROI showing a single particle in the straight- (c) and side-view (f) configurations, in axial steps of ${15}\;\mathrm{\mu} \textrm {m}$. In the latter case, the particles remain in focus upon further insertion of the probe until they exit the FOV.

Download Full Size | PDF

With the side-view probe, we can easily follow the same particles along the whole FOV, with only a slight variation in contrast from one position to the next [Figs. 6(d)–6(f)]. The average intensity projection shown in Fig. 6(e) shows that the fluorescent particles are displaced linearly along parallel trajectories with respect to the fibre probe, with the spacing between their axial positions matching the fibre penetration step of ${5}\;\mathrm{\mu} \textrm {m}$. Moreover, the particles only slightly deviate from the linear trajectory in the $y$ direction, i.e., the direction transverse to the penetration $x$ direction. These are good indications that the medium in the region under examination is not affected by the penetrating fibre probe. Moreover, the linear displacement of the beads along the FOV facilitates tracking their full trajectory by stitching multiple frames.

The fact that each particle keeps a constant contrast, but the contrast varies between particles, indicates that these move at different distances from the focal plane but parallel to it. Again, in Fig. 6(f) we show a single particle in a selected ROI, where between frames the fibre was inserted ${15}\;\mathrm{\mu} \textrm {m}$ deeper in the phantom. The constant contrast evinces that the sample suffers little to no compression, allowing the particles to remain still in the phantom [Fig. 6(f)].

4. Conclusion

This work introduces the first holographic microendoscope based on a single multimode fibre for side-view imaging, a new pathway to reduce footprint of the imaging instrument and further minimize damage to the tissue under investigation. We compared the novel side-view probe and a conventional en face probe, based on identical multimode fibres, in terms of their focusing ability and imaging performance. A higher fraction of controlled power was reached with the side-view imaging probe, which translated into an increased imaging contrast compared to the flat-terminated fibre. Additionally, the side-view imaging probe showed an increased effective NA–accordingly yielding imagery with enhanced spatial resolution–beyond the nominal NA of the multimode fibre used. As confirmed by a simple ray-optics numerical model, this is a consequence of the geometry of the modified fibre termination, whereby the ${5}^{\circ }$-slant reduces the effective core size and promotes the conversion of guided modes to higher-order ones, whereas the ${45}^{\circ }$-angled cleave bends the optical axis by ${90}^{\circ }$ to convey the probe with its side-imaging capability. A reduction of the FOV was seen to accompany the increase in effective NA, which can be understood as a manifestation of the conservation of the optical invariant.

The impact of the fibre probes on the samples was addressed by examining a tissue phantom with dispersed fluorescent microspheres, with a texture mimicking brain tissue. This revealed that, upon insertion of the conventional straight-view imaging probe, the region of the medium being observed undergoes deformation, with the fluorescent particles being centrifugally displaced, and simultaneously dragged axially by the probe in the direction of its penetration in the sample. In contrast, the side-view imaging probe showed negligible impact upon regions of the phantom under observation, with the particles describing linear trajectories and remaining in focus as the FOV moved during fibre insertion. Consequently, this imaging modality facilitates stitching frames together, allowing to create a full map of the sample along the penetration path. Moreover, another dimension can be added to the raster-scanning of the excitation foci by modifying the focal plane, which can be accomplished by multiplying the transmission matrix with a defocus operator [29], therefore allowing volumetric imaging in this configuration. This system exploiting the side-view imaging probe could in principle be upgraded to allow ${360}^{\circ }$ “panoramic” imaging, which would require rotating the sample, or rotating the probe around its longitudinal axis and modifying the sequence of holograms displayed by the DMD for each angular step.