1. Introduction

Wide-field fast optical sectioning of live systems is considered an essential experimental condition for advances in the life sciences [1]. In fact, non-invasive 3D imaging through scattering media, especially biological tissue, is an important and challenging problem in many biomedical applications and a variety of technologies have addressed this challenge [2]. One widely realized technique is confocal scanning microscopy (CSM), which makes use of spatial filtering to provide effective in-depth optical sectioning [3]. Optical coherence tomography (OCT), emerged during the last decade as a state-of-the-art imaging technique. It combines confocal imaging with low-coherence interferometry to extract the unscattered ballistic component of the optical signal [4]. The coherence gate allows effective suppression of multiple-scattered diffuse light. Both, CSM and OCT have relatively long recording times when collecting 3D data due to sequential acquisition of the image pixels. Holographic optical coherence imaging (HOCI) is an alternative depth-resolved optical imaging technique. In this work we use a highly sensitive organic photorefractive (PR) polymer composite as the recording medium. The advantage of holographic recording compared to most other cutting-edge optical imaging techniques, including confocal microscopy and OCT, is the whole-field image formation mechanism, which enables high data parallelism, replacing point-like scanning.

Abramson first suggested that coherence-gated holography might be an effective way of separating image bearing, i.e. unscattered ballistic light, from disrupted multiple-scattered and non-coherent photons when optically probing through biological tissue [5]. The so called ”light-in-flight” recording by holography combines the concept of incoherent time gating using a low temporal coherence light source with holographic techniques resulting in coherence-gated holographic imaging. Abramsons’ approach did not achieve a breakthrough mainly because static holographic film was used as the recording medium, and because the sample was transilluminated, preventing the possibility for depth-resolved imaging. With the development of charge-coupled-device (CCD) arrays and the advent of photorefractive materials, dynamic (reversible) recording media became available for real-time imaging applications. Although CCDs and PR crystals have been known since the 1960s, sufficient sensitivity and practical real-time holographic imaging through scattering media have been achieved only at the beginning of the 1990s [6, 7]. More recently, due to the rapid advances in computational processing power capacities, electronic or digital holography (DH), in which the holographic interference pattern is digitally recorded by a CCD camera, is emerging [8]. DH for depth-selective optical sampling of biological tissue has become feasible using wavelength scanning and coherencedomain imaging [9, 10]. However, DH requires numerical image reconstruction. Unlike DH, PR holography features a purely optical reconstruction process without postdata processing, making the technique a direct imaging approach without computed tomography and free of mathematical artifacts.

Using bulk PR crystals (e.g. BaTiO₃) and semiconductor PR multiple quantum-well devices (PRQW), depth-resolved real-time reconstructions of holographic images have been obtained [11, 12]. Bulk crystals were abandoned quite early because of slow response times (≈1 s) and the difficulty of Bragg-matching volume gratings. PRQWs exhibit high sensitivity and have been successfully adopted for in-depth imaging of biological tissue with high image acquisition speed, approaching frame rates far above video-rate recording [13]. However, PRQW devices require semiconductor device fabrication facilities and have limited spatial resolution (≈10 µm). A promising alternative to this class of materials are organic PR polymer composites. The great chemical flexibility and exceptional long-term stability, combined with the ease with which large thin-film and other device configurations can be realized, are important advantages compared to the inorganic counterparts. There are no material-related constraints with respect to spatial resolution, which makes diffraction-limited transverse resolution possible. Furthermore, PR polymers are amorphous and, therefore, do not introduce additional noise due to optical imperfections as is frequently observed in PRQWs and semiconductor crystals [14]. Moreover, diffraction efficiencies close to 100% are possible [15] compared to 4% in microcavity enhanced PRQW structures [16].

In the past, due to low material sensitivities, holographic imaging with PR polymers has been realized either by transillumination of the sample [17] or using continuous-wave (cw) laser light [18], again, preventing the possibility for depth-resolved imaging. We developed a PR polymer composite with unprecedented high near infrared (NIR) sensitivity and applied it as the recording medium for holographic optical coherence imaging (HOCI). The same material was employed recently to demonstrate the feasibility of using PR polymers as recording medium in coherence-domain holography [19]. Here, holographic imaging was studied using different image-formation systems including image domain (recording device located in the image plane) and Fourier domain (recording device located in the Fourier plane). The need for a temporal coherence gate in addition to the slanting of the PR device with respect to the writing beams (to enable the PR effect) carries with it the problem of beam walk-off [20], which reduces the effective interaction region of the writing beams and, therefore, the field of view of the reconstructed image. We show that for 2D and 3D, image domain holography (IDH) is favorable with respect to image resolution. Moreover, tilting the target may significantly enhance the field of view when imaging specular 2D targets. Therefore, IDH was used to explore the potential of PR polymer composites as rewritable recording medium for coherence-gated holographic imaging of living tissue. Depth-sectioned images of rat osteogenic tumor spheroids as a biological tissue model system are presented. Albeit tumor spheroids have been studied before by HOCI, this had been accomplished either by using PRQW as the recording medium or by employing digital hologram recording [13, 21] and not PR polymers, as presented here.

2. General principle

The imaging of large, biological specimens is challenging because of tissue absorptive and scattering properties. The influence of absorption can be reduced by probing the sample in the so-called therapeutic window, which spans from 600-1300 nm [22]. Image deterioration can be efficiently suppressed by selecting the information-bearing ballistic (unscattered) photons from the background of scattered light. In PR-HOCI this is accomplished by applying coherence-gated holographic recording using a PR device as a coherence filter [14]. The holographic method permits a gate that is as short as the coherence gate delivered by the light source. Consequently, a short coherence length is equivalent to a short pulse (a low-coherence cw source can be used equivalently for coherence-domain holography). This applies to low coherence interferometry, in general. Similar to OCT, HOCI uses backscattered light to enable depth-selective imaging of reflection sites. Thus, light propagates twice through the same object region.

Figure 1 schematizes the principle of coherence gating designed for depth-resolved holographic imaging using a PR polymer device as the recording medium. The image considers a low-coherence probe beam irradiating a turbid medium. The planes within the sample represent heterogeneous features embedded in a surrounding medium. The trajectory of the beam will be dominated by scattering and absorption events, the ratio between both processes depending on the type of medium. Multiple-scattered light loses its coherence after only few scattering events [14]. However, a fraction of the beam will be reflected coherently from the inhomogeneities, forming ballistic photons or ”echoes”. Diffusely-scattered light is typically many orders of magnitude brighter than these echoes. Echoes emerging from the specimen will have a specific time-of-flight (T₁-T₃) corresponding to a fixed depth. In Fig. 1 the echo T₁ exhibits the shortest and T₃ the longest time-of-flight. By combining the signal with a coherent reference, the small coherent echoes can be individually captured from the noncoherent background. A variable delay in the reference arm addresses ballistic light emanating from a specific depth. In PR-HOCI, the holographic reconstructions intrinsically yield transverse sections of the sample, so called en-face images. By scanning the depth with the reference mirror, stacks of transverse HOCI images can be successively recorded, which can be used to construct three-dimensional profiles of the sample.

 

Fig. 1. General principle of coherence-gated holographic imaging using a photorefractive (PR) polymer device as coherence filter.

Download Full Size | PDF

In PR coherence-gated holography, the ballistic and multiply-scattered light emanating from the sample is combined with a reference beam derived from the same light source in a PR medium. However, only backscattered light that retains coherence with the reference, i.e. light path-matched within the coherence length of the source, forms a fringe pattern in the holographic film. In other words, single scattered light detected by the PR device is limited to photons scattered by the coherent probe volume. The coherent probe volume equals the depth of the coherence gate multiplied by the corresponding beam cross section, A. The intensity pattern gives rise to a refractive index modulation Δn (PR grating) within the PR medium. Light that experienced multiple scattering within the sample is presumed to lose coherence and does not contribute to the PR grating. Although the incoherent background causes an erasure mechanism, it generally does not affect the dynamic range of the PR film noticeably. In analogy to OCT that uses temporal heterodyning for signal demodulation, the demodulation of the information stored in the PR film may be described as spatial heterodyning [23]. In holography, the purely optical process of demodulation involves diffraction of the probe beam from the imagemodulated film into the imaging sensor (CCD). The difference between time-dependent OCT and spatial heterodyning in HOCI is the data parallelism. While in temporal heterodyning only one optical mode per scanning event is detected, holography brings into interference many optical modes at a time. The gain in data parallelism in HOCI compared to OCT can be as large as A/λ² where λ is the recording wavelength [23]. HOCI can be compared to wide-field OCT imaging [24], but without the need for image calculation. For comparison, a HOCI system with A=1 mm² and λ=830 nm corresponds to a wide-field OCT system with a detection array of 1000×1000 pixels. Hence, the spatial multiplexing of HOCI may yield an increase in voxel rate as compared to OCT. This demonstrates the potential of HOCI for high-speed image acquisition.

3. Experimental setup

3.1. The HOCI system

Different optical setups have been used for coherence-gated holographic imaging. The Linniktype Michelson and the Mach-Zehnder-type interferometer are the most widely used interferometric arrangements [25]. The first allows a compact alignment of the optical elements but is limited to small angles between the writing beams. It is, therefore, improper for holographic recording using PR polymer devices. Here, we used a slight deviation from the typical Mach-Zehnder interferometer, allowing the angular separation of the writing beams to be adjusted within a large angular range. Figure 2 shows the optical setup with the incorporated PR polymer device. The low-coherence light source was a Kerr-lens mode-locked femtosecond pulsed Ti:Sapphire (Ti:Sa) laser (Clark-MXR, NJA-5, 100 fs pulse duration, 100 MHz repetition rate) with a spectral bandwidth of 37 nm centered at 835 nm and ≈450 mW output power. The laser can be set either to cw or mode-locked operation.

 

Fig. 2. (a) Experimental layout for holographic optical coherence imaging. M: mirror. λ/2: half-wave plate. λ/4: quarter-wave plate. PBS: polarization beam splitter. L1-L3: lenses. PR: photorefractive polymer device. LD: cw laser diode. FP: Fourier plane. The inset shows the positioning of the sample from a side view. (b) Output spectra of the Clark-MXR, NJA-5 femtosecond Ti:Sapphire laser in cw and mode-locked operation. (c) Corresponding fringe pattern (interferogram) of the Ti:Sapphire laser in mode-locked operation using a mirror as target.

Download Full Size | PDF

At the entrance of the interferometer the beam is split into an object and a reference beam. A translation stage was used to control the temporal delay of the reference beam with respect to the light that is coherently backscattered from the investigated object. The samples analyzed with HOCI were arranged such that the ballistic light was backscattered vertically (see Fig. 2, inset). This configuration is necessary because tumors need to remain in culture medium. In the object arm, the combination quarter-wave plate/polarizing beam-splitter (PBS2) ensures that light reflected from the sample is separated from the incoming beam. The lenses L1 and L2 formed the collection optics (1:1 magnification). A Fourier plane is available for spatial filtering through an aperture. This aperture controls the spatial coherence and speckle size at the recording device plane. The optical layout behind the light collecting lenses depended on the image-formation regime. For IDH the image-bearing signal beam was relayed onto the PR device after L2. A third lens was placed between L2 and the PR device for Fourier-domain holography (FDH).

The device formed an off-axis geometry typical for PR polymers with the write beams separated by θ _ext=20°. The angle defined by the bisector of the write beams and the sample normal was 60°, resulting in a refractive index pattern with a period of Λ=4.1 µm. An independent read-out beam counterpropagating against the reference beam was adopted to acquire the holographic images in a four-wave-mixing arrangement. For space-saving reasons, polarizations were chosen ”p” for the write beams and ”s” for the read beam, though this reduces the diffraction efficiency (η) of the PR polymer by a factor of ≈ 3 [26]. A polarizing beam splitter was placed close to the device to optically separate the image-bearing diffracted read beam. A CCD camera (Pulnix, 8-bit) was used to acquire coherent X,Y 2D (en-face) images. Images were background subtracted to account for read beam scattering originating from the PR device.

3.2. The photorefractive recording medium

The ability to respond to coherent light fields is a distinguishing feature of PRs which can be exploited in imaging through turbid media to separate ballistic light from multiple-scattered photons. For the application of PR materials in HOCI, high NIR sensitivity is an indispensable condition. A common expression for the sensitivity S that can be readily determined from experimental quantities is

where t _exp is the exposure time, η _ext the external diffraction efficiency and I _tot,ext the total light intensity impinging on the device. According to Eq. (1), high material sensitivity requires large diffraction efficiency and high recording speed at low light levels. High recording speed is particularly important to prevent image artifacts if the sample is moving or evolving.

The basic principle of photorefractivity requires three functionalities in one material: a sensitizer yielding electron-hole pairs upon illumination, a photoconductor for efficient charge transport (generally holes) and a nonlinear optical (NLO) component, which allows the buildup of an electric-field-dependent refractive index grating via space-charge field formation (linear Pockels effect). In PR polymer composites electro-optical chromophores are typically used to ensure NLO response to the external electric field. The orientational mobility of the chromophores provides an additional contribution to the materials’ dynamic range (orientational enhancement mechanism), which is unique for organic PRs such as PR polymer composites with a low glass-transition temperature (T_g) and PR liquid crystals [27]. Unlike in inorganic PRs, in PR organic guest-host systems the different prerequisites essential for the PR effect are realized by different components. This allows us to address each component individually.

We combined the polymer PF6-triphenylamine dimer (TPD) (40 wt.%), a high-molecularmass hole-transporting polymer consisting of fluorene-bridged TPD moieties, with the sensitizer [6,6]-phenyl C61-butyric acid methyl ester (PCBM, 10 wt.%) and the electro-optical chromophores 2,5-dimethyl-4-(p-nitrophenylazo)anisole (DMNPAA, 25 wt.%) and 3-methoxy-4-(p-nitro-phenylazo)anisole (MNPAA, 25 wt.%) to obtain a high-performance PR polymer composite. The PR material exhibits a T_g of 6 °C. No plasticizer is needed. The PR devices were fabricated by melting the composite between two transparent indium-tin-oxide-coated glass electrodes. The active layer thickness was adjusted to d=105 µm. Figure 3(a) compares the PR performance of the PR polymer presented here to a widely-used PVK-based composite [28]. Overmodulation is observed in the case of the PVK composite for the external electric field applied here. Under similar experimental conditions the recording speed of the PF6-TPD-based material is a factor of ≈250 higher than the PVK-based reference. The PF6-TPD composite exhibits 10% internal diffraction efficiency after 7 ms, applying 0.5 W/cm2 total intensity (beam ratio β=1) of 830 nm cw laser light. This corresponds to a PR sensitivity of 60 cm²/J. In the steady-state regime the PF6-TPD material achieves almost 100% diffraction efficiency. Note that although the steady-state diffraction efficiency may take several seconds to be achieved, it is the early temporal performance of the recording medium which is crucial for the sensitivity of the optical system in HOCI.

 

Fig. 3. (a) Temporal evolution of the diffraction efficiency for the PF6-TPD composite and the PVK-based material PVK:DMNPAA:MNPAA:ECZ:TNFM (41.5:25:25:7.5:1 wt%; λ=830 nm). The external electric field was turned on 30 s before hologram recording was started (pre-poling step). (b) Comparison of the steady-state performance for the PF6-TPD material using a cw laser diode (LD, λ=830 nm), a superluminescent diode (SLD, λ=840 nm, l_c=42 µm) and a femtosecond pulsed Ti:Sapphire laser (Ti:Sa, λ=840 nm, l_c=8 µm) for hologram recording. The optical polarization was ”s” for the write beams and ”p” for the read-out beam.

Download Full Size | PDF

The exceptionally high sensitivity of this new material is mainly due to the relatively high hole mobility of the PF6-TPD polymer as compared to PVK and other polymers previously used in PR polymer composites and the relatively high amount of sensitizer [29]. The 10 wt.% PCBM that were incorporated into the composite represent the optimum sensitizer concentration for the present material with respect to the recording speed. No pre-illumination of the material is required. Overall, the sensitivity of the PR composite reported here represents an improvement by a factor of three compared to the best PR polymer device reported before [30].

4. Resolution and field of view

4.1. Lateral and axial resolution

Unlike transillumination holography, where only the transverse (lateral) resolution can be specified, HOCI and low coherence interferometry (LCI) in general permit optical depth-sectioning of volumetric targets with an accuracy that is defined by the longitudinal (axial) resolution. In contrast to other optical imaging techniques, e.g. conventional and confocal microscopy, the longitudinal resolution in LCI is independent of any focusing conditions. Moreover, lateral and axial resolution do not depend on each other.

In HOCI, the lateral resolution depends on the holographic imaging regime. In IDH, the spatial resolution depends mainly on the resolving capacity of the recording medium, while in FDH the active window size of the device is normally the resolution limiting element. The Rayleigh criterion may be used to define the transverse resolution, which reads

R_s is the minimum resolvable distance between two points, λ the wavelength, f the focal length of the imaging lens and M is the magnification of the optical system. In FDH, D defines the width of the device (clear device aperture), whereas in IDH D corresponds to the aperture of the imaging lens [10]. For comparison, in conventional microscopy and OCT the transverse resolution for imaging is determined by the focused transverse spot size of the optical beam, given by Δx=(4λ/π)(f/d) where d is the spot size on the objective lens and f is its focal length. The transverse resolution of our optical system as calculated by the Rayleigh criterion was 20.4 µm in the FDH regime (f=50 mm, D=2.5 mm) and 8.0 µm in the IDH regime (f=200 mm, D=25.4 mm).

In analogy to OCT, the axial resolution in HOCI is mainly governed by the coherence length l_c of the light source. l_c is the spatial width of the field autocorrelation as measured by the interferometer. The envelope of the field correlation however is equivalent to the Fourier transform of the power spectrum of the source (Wiener-Khinchin theorem). Thus, the axial resolution in LCI is inversely proportional to the spectral bandwidth of the light source (in confocal microscopy the axial resolution is determined by the depth of field). The relation between the coherence time or length and the spectral width depends on the source spectral profile. Assuming a Gaussian spectrum and considering further that light travels twice through the sample it can be shown that

where Δz is the so-called round trip coherence length and a measure for the depth resolution in LCI when imaging translucent samples [31]. Here, l _FWHM is the full-width-half-maximum (FWHM)-width of the interference fringe signal (interferogram), λ is the central wavelength and Δλ is the FWHM-width of the spectrum in wavelengths. Eq. (3) is generally used in OCT to predict the axial resolution of the optical system and can be equivalently applied to HOCI. Figure 2(b) shows the emission spectra of the Ti:Sa laser used here in both modes of operation. Applying Eq. (3) to calculate the depth resolution from the power spectrum in pulsed operation gives 8.3 µm. Figure 2(c) shows an interferogram of the Ti:Sa laser, measured using a high-reflecting mirror at the sample position. The interference fringe signal reveals a FHWM-width of 24 µm. This corresponds to a round trip coherence length or axial depth resolution of l _FWHM/2=12 µm. The discrepancy from the theoretical value is most likely due to the absence of dispersion compensating elements and the use of dielectric mirrors in the optical layout.

4.2. The beam walk-off effect

The typical arrangement of the writing beams and the polymer device for holographic recording was previously developed in order to allow maximum space-charge field formation and hence refractive index modulation [15]. This is important when high system sensitivity is envisioned. The standard transmission geometry as used here, however, may lead to constraints when performing coherence-gated holographic imaging with low-coherence light sources. In the limiting case of large beam diameters, i.e. when the coherence length is less than the beam diameter at the recording medium (which is generally the case), the angle between the writing beams causes the effective interaction region to become smaller than the region of the beam overlap [32]. This effect is called ”beam walk-off”. The problem has been addressed for digital holography and for holographic recording using either BaTiO3 or PRQW devices as recording medium, but not for PR polymer devices [20, 33, 10]. When using PR polymers as the recording medium, the impact of the walk-off effect is even more pronounced because of the relatively large angle of intersection (20° compared to ≤2° in the media referred before). The influence of a reduced interaction region on the PR performance of the device is illustrated in Fig. 3(b). The PF6-TPD-based material was probed under similar recording conditions using a cw laser diode (λ=830 nm), a superluminescent diode (SLD, λ=840 nm, l_c=42 µm) and a femtosecond pulsed Ti:Sapphire laser (Femtolasers, Fusion PRO-800, 75 MHz, λ=840 nm, l_c=8 µm). Due to the high repetition rate, the femtosecond pulsed light acts on the PR medium as quasi-cw light. For this reason, the normalized grating buildup time under low coherence excitation is roughly the same as when a cw laser under identical conditions is used [19]. The diffraction efficiency η _int, however, and thus the refractive index modulation amplitude Δn is reduced because of the problem of beam walk-off. Note that η _int decreases by the same factor as does the coherence length and, therefore, the effective intersection region when going from the SLD (η _int=0.16) to the Ti:Sa laser (η _int=0.03).

 

Fig. 4. Holographic imaging regimes and their influence on the field of view and image resolution using a PR polymer device as the recording medium: Fourier-domain holography FDH (top) and image-domain holography IDH (bottom). In FDH the field of view is preserved but the resolution is severely affected in the horizontal direction of the reconstructed image. In IDH the resolution remains intact, the field of view being restricted by the effective interaction region. (a) and (b) show holographic images of the USAF target and of diffuse paper with the symbol ®, respectively, in mode-locked operation using FDH. (c), (d) and (e) show the same targets using IDH.

Download Full Size | PDF

Additionally, the limited interaction region in the intersection area of the write beams influences which holographic image formation mechanism is to be implemented in the HOCI system, as schematically illustrated in Fig. 4. Propagation of the pulses, or coherence gate, inside the medium limits the interaction region to a stripe [19]. The width of the effective interaction region w in the large beam case can be derived from geometric considerations and is

where θ _int=6.8° is the internal angle between the write beams. In our arrangement this results in a field of view of 182 µm (for a refractive index of 1.7). FDH is usually more favorable than IDH by preserving the field of view in FDH and because of its reduced sensitivity to background scatter originating from the holographic film [34]. However, the effective interaction region will act as a spatial filter and cut off part of the high-frequency components when the recording device is located in the Fourier plane. More precisely, the vertically elongated and horizontally restricted hologram width will accommodate the vertical frequencies and disregard part of the horizontal frequency components of the Fourier spectrum. If, however, the recording device lies in the image plane, this will simply reduce the field of view of the hologram to the width of the effective interaction region. These considerations were verified experimentally by imaging an U.S. Air Force (USAF) test target and diffuse paper. Figure 4 compares the hologram reconstructions as expected from the analysis made before to the experimental results. Only hologram reconstructions of relatively poor resolution were obtained in the FDH regime (Fig. 4, top). As is evident from the USAF reconstruction [Fig. 4(a)], horizontal bars are better resolved than vertical bars. This is because in the horizontal direction the transverse resolution (≈400 µm) is limited by w while in the vertical direction (≈67 µm) the limit is imposed by the vertical width of the polymer device (2.5 mm). When imaging diffuse targets using FDH [Fig. 4(b), black ® on paper] the hologram reconstructions were dominated by horizontal elements (stripes), showing poor resolution in the horizontal direction.

Figures 4(c, d) and 4(e) show reconstructed holograms of the USAF test target and the printed ® on diffuse paper, respectively, using IDH. In this case, the transverse resolution remains unaffected, the field of view being restricted and defined by the effective interaction region. From Fig. 4(d) the transverse resolution can be measured to be 9.8 µm, which is close to the theoretical limit as imposed by Eq. (2). No restriction to the field of view occurs when a cw light source is used (not shown, see Ref. [19]). In this case, the image cross-section is defined by the beam diameter on the target.

Figure 4 assumed that the object wavefront is perpendicular to the light propagation direction. This is not necessarily the case and, in fact, the walk-off effect can be suppressed by tilting the target, as is shown in Fig. 5(a). If the time-of-flight (TOF) of the object wavefront is tilted relative to the wavefront of the reference beam, it is possible to match the coherent wavefronts within the recording medium and increase the field of view [10].

Figure 5(b) presents the typical off-axis hologram recording geometry for PR polymer devices. The scheme is used to calculate the tilt angle γ that allows the wavefronts to be matched within the PR polymer. The tilt angle γ can be expressed as

To calculate the tilt it is necessary to evaluate the angles α and β, which are given by

The factor ξ corrects for the change of the beam diameter due to refraction. ϑ₁=70° and ϑ₂=50° are the external angles with respect to the device normal for the reference beam and the signal beam, respectively. n is the index of refraction. Substituting Eq. (6) and Eq. (7) into Eq. (5) gives γ=8.0°. This is the target tilt necessary to match the coherent wavefronts within the PR polymer device. However, the result needs to be considered with some care, since it may be invalidated by several factors. First, the light backscattered from the target before striking the device runs through several optical elements. The optics may not support the angle necessary to compensate the walk-off effect, e.g. due to aberration and vignetting effects. On the other hand, matching the wavefronts means that the internal angle of the write beams is reduced to a minimum, and consequently the fringe period Λ will become very large. A large fringe period, however, generally dictates a loss in dynamic range and a reduction of recording speed. Both of these adversely affect the sensitivity of the system. If the fringe period becomes larger than the mean free path of the charges, no PR grating will be produced at all. Furthermore, a strong tilt of the object results in out-of-focus compared to the non-tilted situation by increasing the distance of the parts of the object to the focal plane [10]. Hence, there is a twofold trade-off between field of view and resolution, on the one hand, and between field of view and sensitivity on the other hand. In our experiments, the target was tilted to accommodate a larger field of view. By doing so, an optimized field of view in the order of 1.3 mm was achieved [Fig. 4(d) and Ref. [19]]. However, the field of view was not fully recovered because our optical system could not support the angle necessary to completely compensate for the walk-off in the PR polymer. Loss of ballistic light due to vignetting forced us to choose a sub-optimum angle.

 

Fig. 5. Geometric considerations for beam walk-off elimination in a PR polymer device. (a) The walk-off is intrinsically suppressed for a 3D object. TOF: time-of-flight. Adapted from Ref. [10]. (b) Typical off-axis recording scheme for PR polymers. For a 2D target the beam walk-off effect can be suppressed by tilting the target.

Download Full Size | PDF

Although not explicitly mentioned, the considerations from above referred to the situation when a 2D target is imaged. In a 3D target there is not one but a distribution of TOFs, as denoted in Fig. 5(a). When scanning the delay line, the reference wavefront will always match a complementing wavefront with a fitting TOF. Therefore, holograms from translucent volumetric targets intrinsically suppress the walk-off effect, provided that the recording medium supports large fringe periods.

 

Fig. 6. Coherence-gated holographic images of test targets and tissue samples: (a) polyurethane foam [19], (b) unstained human skin section (60 µm) in which two hair follicles are discernible (inset: ×20 microscope image of an adjacent stained slab. The region shown in the holographic image is outlined), (c) 73 µm thick trabecular bone sample, imbedded in a transparent matrix for fixation. The lamellar layers below the ”matrix” arrow and above the ”bone tissue” arrow are due to thickness variations inside the matrix.

Download Full Size | PDF

5. HOCI of living tissue

To demonstrate the feasibility of using PR polymer composites for coherence-gated holographic recording, we applied HOCI to optically section various volumetric test targets and biological samples. Figure 6 shows hologram reconstructions of a polyurethane foam [Fig. 6(a)], which is used as a cell culture matrix in regenerative medical applications and was studied elsewhere [19], a thin human normal skin slab [Fig. 6(b)] and a human trabecular bone section [Fig. 6(c)]. Figure 6(b) and 6(c) were too thin to effectively achieve optical depth-sectioning with the laser system used here. Therefore, rat tumor spheroids were studied. Tumor spheroids are a wellknown in vitro tissue model system that are relatively easy to handle, e.g., for the systematic study of tumor response to therapy [35]. The tumor spheroids were grown from rat osteogenic sarcoma cells, so called UMR-106 cells. For HOCI, tumors with ≈1 mm diameter are chosen. The cell culture process is described in detail in Ref. [23]. Agar is used to solidify the culture medium and at the same time fix the tumors at the bottom of a petri dish. Details about the anatomy and physiology as well as the optical properties of rat osteogenic sarcoma spheroids can be found in Refs. [36, 37, 38]. For the discussion in this work, it is important to reiterate that in such spheroids the cells in the peripheral region are rapidly dividing, whereas deeper in the spheroid there is negligible cell division because of nutrients and oxygen insufficiency. After sufficient growth, cell apoptosis and necrosis occurs in a manner similar to naturally occurring spherical tumors. Therefore, the tumors consist of an outer shell with relatively high cell motility and of a necrotic core. The necrotic tissue exhibits different optical characteristics from the healthy outer tissue. In addition, microcalcifications may be developed towards the center of the tumors.

Figure 7(a) shows a cw hologram of a 740-µm-diameter tumor in a false color image, in which the reflectivity increases towards the red end of the scale. Although depth-selective imaging is not possible under cw conditions, the image reveals the general morphology of the tumors. The tumors have a spherical symmetry and are made of sub-resolution structural features. Therefore, the holograms consist of specklelike features that result from those sub-resolution elements. In the cw image, the outermost shell can be distinguished from the inner core. The shell is approximately 100 µm thick. Holograms recorded in cw operation intrinsically reveal signals from the surface as well as from the inside of the tissue. It is not possible to allocate the origin of the signals. On the other hand, by operating the Ti:Sa laser in mode-locked condition 2D cross-sections from inside the tumor tissue can be selectively recorded. In the present case, the holographic OCI volumetric data set consisted of 100 successively recorded X,Y frames, measured with a step delay of typically 25 µm. The tumors were irradiated with ≈ 5 W/cm². The total recording intensity at the PR device was 5.5 mW/cm² with a reference-to-signal beam ratio of ≈1. The acquisition time was 500 ms/frame. The electric field (dc) applied on the PR composite was 57 V/µm. Continuous read-out of the hologram with 0.5 mW/cm² of an 830 nm emitting cw laser diode yielded 1.3×10^-3 external diffraction efficiency. The achievable rate of the present system can be calculated to be 4×10⁵ pixels per frame (given by the CCD) ×100 frames per fly-through=4×10⁷ voxels (corresponding to a volume of ≈1×1×2.5 mm³) in a 50 s total integration time. This yields a voxel rate of 8×10⁵/s which is in the same order as the values reported before using PRQW and DH [23, 10]. Note that the field of view was sufficiently large to acquire the full shape of the tumors.

 

Fig. 7. Selected views of a 740-µm-diameter rat osteogenic sarcoma tumor spheroid. (a) Holographic image in cw operation. (b) Integrated view of the spheroid produced by summation of the en face sections of the full data set. (b) Medial en face cross-section. Each image was scaled separately.

Download Full Size | PDF

Figure 8 shows selected X,Y (en face) sections from the volumetric data set taken from the same 740-µm-diameter tumor as shown in Fig. 7(a). A dynamic range of approximately 20 dB was achieved. The bottom of the tumor is seen in frame 8. The middle of the tumor is approximately at frame 22 (350 µm or 700 µm total path). At this depth, the hologram is strong, but becomes weaker for deeper sections and disappears around frame 33 (≈ 600 µm or 1.2 mm of total path). The optical sections from inside the tumor reveal an intensity pattern that results from tissue morphology variations. Towards the center of the tumor the reflectivity increases because of the occurrence of necrosis and microcalcifcation, as outlined before. This is particularly clear up to frame 18, where the backscattered signal is still strong. These findings are in accordance with previous results on rat osteogenic sarcoma tumors obtained with HOCI [23].

Figure 7(b) shows a pseudo-transillumination image of the same tumor, generated by integrating over the whole en face data set of Fig. 8. Figure 7(c) is an enlargement of a medial X,Y cross-section of the spheroid. Each image in Fig. 7 was scaled individually. Similar to the cw image [Fig. 7(a)], the pseudo-transillumination image clearly distinguishes between the outer shell (low reflectivity) and the core (high reflectivity) of the tumor.

 

Fig. 8. Selected X,Y (en face) sections of a 740-µm-diameter tumor spheroid immersed in growth medium. The holograms were captured with a step delay of 25µm. The bottom of the tumor is at frame 8. The images are displayed in a logarithmic intensity scale.

Download Full Size | PDF

 

Fig. 9. Volumetric visualizations of rat tumor spheroid (∅≈800 µm). (a) Collection of 350 pseudo-A-scans selected from stacked en face holographic images. (b) 3D rendering in a box generated from a fly-through data set. (c) 3D rendering of the same tumor with a vertical cut. The light was incident from the top.

Download Full Size | PDF

We succeeded in imaging the entire volume of an approximately 800-µm-diameter tumor spheroid. However, in this particular measurement the dynamic range was only 10 dB. Figure 9(a) shows a collection of 350 so called pseudo-A scans in a dB scale, i.e. reflectivity versus depth along selected Z-lines, extracted from the center area (100 µm×100 µm) of the holographic en face images. The graph shows that the reflectivity increases first to an approximate depth of 200 µm (healthy shell) and then decreases exponentially. The shape of the pseudo-A-scans is consistent with measurements presented by Jeong et al. using digital HOCI [10]. A penetration depth of ≈800 µm within biological tissue at a relative low dynamic range as observed here suggests that multiply-scattered coherent light from the extreme depths is captured by the coherence gate. This is an aspect that is currently under investigation.

Figure 9(b) and 9(c) show 3D reconstructions of the tumor spheroid, computed from the volumetric data set. The cut-away volume enhances the 3D rendering of the sample [Fig. 9(c)]. The reason for the flat tumor bottom is the relatively low sensitivity of the system.

For comparison, in previous investigations on tumor spheroids using HOCI Jeong et al. reported a dynamic range of 40 dB and penetration depths of 800 µm when imaging tumor spheroids with holographic OCI using PRQW for hologram recording [39]. In digital HOCI, penetration depths of 1.4 mm for the rat tumors along with a dynamic range of 25 dB were demonstrated [10].

6. Conclusions

The results presented here show the first demonstration of coherence-gated holographic imaging of living tissue using a highly sensitive PR polymer composite as the recording medium. The fabrication of PR polymer devices is very straightforward and free of optical scattering centers. Moreover, image retrieval is a purely optical process and does not require any sort of computation. These are important advantages compared to DH which make PR polymers attractive candidates for HOCI.

We found that for the recording geometry required for this class of materials Fourierdomain hologram recording using low-coherence light strongly reduces the horizontal resolution. Therefore, image-domain holography was implemented in the optical system. This limits the field of view but preserves the resolution. Furthermore, we demonstrated that in the case of 2D targets, tilting significantly increases the field of view of the hologram, while in the case of translucent volumetric samples the field of view is only slightly affected. Based on our results, we may anticipate that using PR polymer composites as recording media in the standard transmission geometry, will not allow us to completely eliminate the beam walk-off effect. For this reason, measurements are in progress in which the polymer device is arranged in a reflection geometry instead.

Motivated by the feasibility of ’optical biopsy’ we showed in-situ optical depth sectioning of rat osteogenic sarcoma tumor spheroids. The sensitivity of the optical system including the PR polymer device permitted us to detect coherent signals to a depth of ≈800 µm with a data acquisition rate of 8×10⁵ voxels/s. This allowed us to clearly distinguish the outer shell and the inner necrotic core of the tumors within a depth range of ≈600 µm. However, improvement of the present design of the setup to incorporate the ideal recording polarizations of the PR polymer in combination with a CCD camera with better NIR quantum efficiency and dynamic range will allow for significantly higher system sensitivity and voxel rate.