Reflection confocal microscopy (RCM) is an emerging clinical cross-sectional method for imaging scattering specimens [1,2]. RCM exploits endogenous scattering contrast that is natively present in biological specimens. RCM employs geometric principles to achieve three-dimensional localization of light that is backscattered from a sample [3]. The specific geometric principle is pinhole illumination with subsequent backscatter detection through an identical or similar pinhole aperture that is proximal to a photodetector. As typically implemented, RCM is a point-scanning modality. That is, in addition to axial scanning to acquire pixels parallel to the direction of light propagation, two-dimensional en face scanning is required to acquire pixels orthogonal to the optical axis. Thus, RCM employs serial acquisition of pixels that requires beam scanning along each axis, along which image pixels are acquired.

Line-field RCM (LF-RCM) is one approach to increasing the degree of parallelization in RCM and to reducing scanner burden [4–9]. A typical approach to realizing LF-RCM is to shape a single-mode laser beam into a one-dimensional line (slit) profile using a cylindrical lens. In this approach, the degree of confocality in the en face plane is explicitly reduced. That is, to afford a degree of parallelization, one degree of confocality is sacrificed in going from a point (pinhole) source to a slit source. When read out at a slit detector, slit illumination enables en face geometric confocality perpendicular to the slit, but is non-confocal along the slit itself. This lack of confocality along the slit raises the possibility that coherent crosstalk could emerge within the slit when imaging a scattering sample. This coherent crosstalk would be expected to be present in addition to, and in contradistinction to, speckle. For definitional clarity, we refer to speckle when discussing coherent addition of random field phasors within a resolution element and to coherent crosstalk when discussing coherent addition of random field phasors that arise from a multiplicity of resolution elements.

Beyond the basic questions related to coherent crosstalk in line-field confocal microscopy, when using a single-mode source, new technological capabilities raise the possibility of designing new sources that mitigate crosstalk. In particular, the advent of low spatial coherence (so-called “speckle-free”) lasers enables the generation of laser light with custom coherence properties. Indeed, coherence engineering has been demonstrated in random lasers [10,11], chaotic cavity lasers [12], vertical-cavity surface-emitting laser (VCSEL) arrays [13,14], and degenerate lasers [15,16]. By virtue of their reduced spatial coherence, these sources may offer performance enhancement in LF-RCM.

In this Letter, we demonstrate a LF-RCM system that uses a reduced spatial coherence line source. The reduced spatial coherence line source is synthesized using a dense VCSEL array. In the absence of scattering, the confocal performance of the reduced spatial coherence LF-RCM system is similar to a spatially coherent LF-RCM system. In the presence of nontrivial sample scattering, however, coherent crosstalk significantly degrades en face images generated with spatially coherent light. On the other hand, en face image quality is notably improved by using reduced spatial coherence light. Nevertheless, the axial confocal performance is significantly degraded with both fully coherent and reduced coherence light. We additionally propose a model to support these results. Our model states that within-line scattering leads to an increased background (non-confocal) signal in the setting of scattering. That is, within line, the imaging system is essentially a traditional microscope that lacks cross-sectioning capabilities. In the case of single-mode light, the background light can coherently interfere, leading to coherent crosstalk. This crosstalk imparts features to the image that are not in the object itself. In the case of reduced coherence light, the background is nominally incoherent and, consequently, relatively uniform. With either type of illumination, however, the presence of non-confocal background generated by a scattering object impairs one of the characteristic features of a confocal microscope: a sharply attenuating total integrated image signal with defocus.

Our LF-RCM system (Fig. 1) has similarities to prior line-field systems [9,17]. The general operating principle is that a line source is imaged onto a sample and that backscattered light is imaged through a detector-side slit before being read out by an array detector. Here, we use a high-speed line scan camera as both the detector-side slit and as the array detector. In this system, the line source is synthesized from a subset of emitters in a dense VCSEL array of narrowband emitters ($\lambda =858\mathrm{nm}$, Princeton Optronics). The source array itself is imaged onto the aperture stop (AS), which is composed of two orthogonal slit apertures, through lens L1 (50 mm in focal length) and L2 (150 mm in focal length) with a magnification of 3. The first aperture is used to control the number of rows of VCSELs that will illuminate the sample. The second aperture is employed to control the number of columns of VCSELs in use. Each VCSEL at the AS is condensed by the cylindrical lens CL1 (50 mm in focal length) into a line along the $x$ direction at the field stop FS1, which is an adjustable slit aperture to control the width of the line focus on the sample. The sample is scanned along the $y$ direction by use of a galvanometer scanning mirror (GSM). In this design, the VCSELs along the same row or the $x$-direction will be angularly compounded at the field stop FS1 as enhanced lines, while the condensed lines from VCSELs at different $y$ positions or rows will be shifted along the $y$-direction. To determine how many rows of VCSELs will contribute to the sample illumination, we gradually increase the opening of the first aperture of AS until the image intensity stop increasing. This way, we determined that approximately five rows of the VCSELs will contribute, as shown in Fig. 2(a). The net effect is one of angular compounding; that is, a large number of line sources, each generated from an independent oscillator, illuminate the sample at a range of incident angles. In the experiment, we set the slit width along the $y$ direction of the field stop FS1 to be $\sim 100\mathrm{\mu m}$, which is imaged onto the sample through lens L4 (200 mm in focal length), and an objective lens (NA 0.15; focal length, 20 mm) by a demagnification of 10. We found further reduction in the slit width of FS1 will lead to a decrease in image intensity. This means that the diffraction effect from the slit begins to affect imaging. The line-scan CCD camera (14 μm along $x$ by 28 μm along $y$; e2v AVIIVA EM2, model: EV71YEM2CL1014-BA9) is put at the imaging plane of the sample through the object lens MO, lens L5 (focal length, 400 mm), lens L6 (60 mm), and lens L7 (150 mm), with a magnification of 50. For comparison, the system is configured so that a traditional single-mode source (i.e., a HeNe laser) can be used for spatially coherent light source illumination.

 

Fig. 1. Diagram of the imaging system. L1–L2, regular lenses with focal lengths of 50 mm and 150 mm, respectively; AS, aperture stop with two orthogonal slits to control the number of VCSELs; BS1–BS2, beam splitters; L3, regular lens of a focal length 100 mm, respectively; CL1–CL2, cylindrical lenses with focal lengths of 50 and 100 mm; FS1–FS2, slit apertures as field stops; L4, regular lens with a focal length of 200 mm;. GSM, galvanometer scanning mirror; MO, objective lens with a NA of 0.15 and a focal length of 20 mm; S, sample. L5–L7, regular lens with focal lengths of 400, 60, and 150 mm; FM, flipping mirror to port a HeNe laser into the imaging system.

Download Full Size | PDF

 

Fig. 2. Images of 1951 US Air Force Resolution Test Chart under different light sources. (a) Image of the light source with $\sim 115\mathrm{V}$ CSELs. (b) Image of the test chart under the light source (a). (c) Image of the light source with $\sim 15\mathrm{V}$ CSELs. (d) Image of the test chart illuminated by the light source (c). Line intensity plot of the edge feature denoted with the red line in (b), (d), and (f). (f) Image of the test chart illuminated by a HeNe laser. We attribute the presence of the intensity variation in (f) to diffraction caused by propagation through the optical system and the sample.

Download Full Size | PDF

Using this LF-RCM microscope, we first imaged a non-scattering object using illumination of different degrees of spatial coherence. The confocal images of the United States Air Force Test Chart (US-AFTC) were similar when using light from 23 columns and when using three columns of the VCSEL array. When using a HeNe source, there is notable variation in image intensity that can be attributed to a diffraction artifact in the setting of high spatial coherence.

We next imaged a Teflon film (Fig. 3) that has some patterned structure that is evident on white-light microscopy. Although the LF-RCM image is speckled when using light from 3 and 23 columns, there is evidence of the patterned structure. However, when using HeNe illumination, the pattern is not evident. We attribute the loss of the pattern to coherent crosstalk. The crosstalk has a speckled appearance that scrambles the information content that otherwise would be present in the image.

 

Fig. 3. Images of a Teflon film. (a) Image with 115 VCSELs. (b) Image with 15 VCSELs. (c) Image with a HeNe laser. (d) Image with a white-light bulb. The white-light micrograph indicates inhomogeneous roughness to the Teflon film. This inhomogeneity is evident under LF-RCM with partially coherent illumination (a) and (b). However, LF-RCM with a fully coherent illumination yields a homogenous speckle pattern that obscures the subtle inhomogeneous detail present in the film.

Download Full Size | PDF

To further support this perspective, we then imaged an US-AFTC with its features placed against the highly scattering Teflon film. That is, following the direction of propagation, the illumination light goes from air to glass to the test chart (with reflective feature) to the scattering Teflon film. The position of the scattering film juxtaposes high scattering against well-defined, high- contrast features. The images in Fig. 4 demonstrate the key aspects of using high versus partial spatial coherence light in direct-detection, line-field confocal microscopy in the presence of scattering. The presence of speckle persists, even when reduced spatial coherence light is used. Speckle is present because the slit acts as a 1D pinhole, creating spatial coherence along the short axis of the slit. That is, there is some degree of coherent summation of randomly phased fields within individual resolution elements. However, reduced or partial spatial coherence limits the degree of coherent summation of randomly phased fields between neighboring resolution elements. This coherent summation between neighbors is a more formal statement of what is meant by crosstalk. As a consequence, the features that are distinctly present when using reduced spatial coherence illumination are difficult to identify or unidentifiable when using highly spatially coherent light. In the experiment, to eliminate strong specular reflection from the glass substrate, the sample was slightly tilted, which can explain the difference in the appearances of the vertical and horizontal features.

 

Fig. 4. Imaging an artificial sample where a Teflon film is tightly pressed onto a negative resolution test chart. (a) Diagram of the light propagation through and reflection by this sample. The shaded regions represent chrome. (b) Image with 115 VCSELs. (c) Image with 15 VCSELs. (d) Image with a HeNe laser.

Download Full Size | PDF

A schematic of our model of coherent crosstalk is shown in Fig. 5. Again, in the setting of line-field illumination and detection, there is some degree of speckle. For reduced spatial coherence illumination, improperly mapped photons within the line add incoherently to increase the background upon which the confocal signal sits. However, for high spatial coherence illumination, the improperly mapped photons are coherent with respect to each other, leading to coherent crosstalk.

 

Fig. 5. Model of coherent crosstalk when the light of high and partial spatial coherence is used. Black (gray) lines are properly (improperly) mapped photons. The dashed lines are field phasors that add coherently.

Download Full Size | PDF

Lastly, we explored the axial response of the LF-RCM. As mentioned above, the imaging system acts as a conventional microscope within the confocal slit itself. Moreover, as noted in Fig. 5, improperly mapped photons generate a background intensity level that does not contribute to imaging information. This background intensity would be expected to degrade the axial performance of the line-field confocal system in the setting of scattering. Indeed, as shown in Fig. 6, the integrated image intensity decreases significantly less slowly with defocus in the setting of scattering for both coherent and partially coherent illumination. We also note that coherent (interferometric) detection of the backscattered partially coherent field would be expected to mitigate the coherent crosstalk that leads to this degradation in axial sectioning performance [18,19]. That is, when the sample is illuminated with partially coherent light, the improperly mapped photon signal would be expected to be manifest in the non-interferometric signal and not the confocal interferometric signal. Implementing coherent detection for the line-field configuration demonstrated in this Letter is a focus of future work.

 

Fig. 6. Axial responses of LF-RCM for a mirror and Teflon film illuminated by different light sources. (a)–(c) Axial responses for mirror target illuminated by 115 VCSELs, 15 VCSELs, and a HeNe laser. (d)–(f) Axial responses for Teflon film illuminated by 115 VCSELs, 15 VCSELs, and a HeNe laser. We attribute the asymmetry of the axial responses to the asymmetric light distribution outside the focal plane due to an astigmatism introduced by the cylindrical lens [8,17].

Download Full Size | PDF

In summary, our results highlight the importance of spatial coherence engineering in parallelized reflection confocal microscopy. We investigated the impact of partial spatial coherence illumination on the performance of line-field confocal microscopy. We note that partial coherence improves the performance in the setting of scattering compared to high spatial coherence. In particular, for high spatial coherence illumination, improperly mapped photons are mutually coherent, leading to coherent crosstalk between adjacent resolution elements. This crosstalk is mitigated by partial coherence. However, regardless of the degree of spatial coherence, scattering degrades axial resolution, which can be attributed to a lack of confocality within the slit itself for the direct-detection imaging system presented. We lastly note that we used a lower NA objective lens in this work ($\mathrm{NA}=0.15$) and that an interesting area of future research is to investigate the relationship between the degree of partial coherence and a coherent artifact as a function of NA.