A novel confocal microscope designed for use with turbid media is proposed. We use a Zeeman laser as the light source. Based on the properties of two-frequency polarized photon-pairs and the common-path feature of polarized photon-pairs with heterodyne detection employed in the proposed confocal microscope, three gatings (spatial filtering gating, polarization gating and spatial coherence gating) are thus simultaneously incorporated in the microscope. Experimental results for the angular distribution of polarized photon-pairs in a scattering medium indicate that polarization gating and spatial coherence gating preclude the detection of multiply scattered photons, whereas the pinhole selects the least scattered photon-pairs. Thus, better performance for axial resolution than can be obtained with a conventional confocal microscope is demonstrated experimentally. In addition, the proposed microscope is able to either look deeper into a turbid medium or work with a denser medium; furthermore, the axial resolution is improved.
Looking into a scattering or turbid medium always appeals to scientists, because the process has many applications, especially in biomedical imaging (Anderson et al., 1994). Although confocal microscopy yields good tomographic images of an object, it does not work well for objects in a turbid medium, because multiply scattered photons arise in the scattering medium. As a result, they overwhelm the weakly scattered photons or snake photons (Chou et al., 2000) that carry the original object information, including the polarization and the direction of propagation. It has been found that the use of a heterodyne technique, in which the signal light from the object through the scattering medium is mixed with a reference light moving along a different path with an interferometric geometry (Kempe et al., 1997), results in the signal that is produced by weakly scattered photons being enhanced. In the heterodyne microscope, the signal is proportional to the product of the amplitude of the image and of the reference beam; this provides optical amplification and enhancement of the signal-to-noise ratio (SNR) of the heterodyne signal. Thus, the working range in a scattering medium of optical thickness τ = µs × l can be improved 11–14-fold in terms of mean free path (Kempe & Rudolph, 1996), where µs is the scattering coefficient of the scattering medium and l is the thickness of the medium. Kempe et al. also introduced the linear correlation microscope (Kempe & Rudolph, 1994) by employing a broadband light source in a heterodyne microscope; through activation of temporal coherence gating, there was an improvement in the axial resolution through thick layers. Similarly, Izatt et al. (1994) incorporated a single-mode optical fibre acting as a pinhole in an optical coherence tomographic system, to successfully make it an optical coherence microscope for confocal imaging. Sheppard et al. (2004) have compared the image formation performance of low-coherence interference microscopes and an optical coherence tomographic system using a high numerical aperture objective. Based on the above, these three microscopes activate temporal coherence gating, which plays a dominant role in imaging sectioning in a scattering medium.
Gan et al. (1999) studied polarization gating to improve the resolution of microscopic imaging through turbid media. Their experimental results showed that polarization gating is able to suppress scattered photons, and thus enhance the image. Peng et al. (2001) proposed a novel Zeeman laser scanning confocal microscope (ZLSCM) for sectioning imaging in a scattering medium with correlated polarized photon-pairs (PPPs). They travel in a common path in the scattering medium, resulting in more scattered PPPs contributing to the heterodyne signal. Thus, the SNR of the detected signal is greatly enhanced because the polarization gating and the spatial coherence gating are activated simultaneously. This is in contrast to the conventional confocal microscope (CCM) based on single-photon propagation (Schmitt et al., 1994), where only the pinhole aperture suppresses the multiply scattered photons in the scattering medium. In order to understand the propagation behaviour of PPPs in the scattering medium, angular distributions of scattered PPPs at different concentrations and with different pinhole sizes were measured. Wax et al. (2001), using low-coherence interferometry, measured the angular distribution of backscattered light in a scattering medium by adjusting the propagation angle of the reference beam for heterodyne detection. Their experimental results indicated that the angular distribution depends upon the tilting angle of the reference beam from the object beam, and the heterodyne signal decreases as the tilting angle is increased. Thus, the behaviour of propagation of PPPs in terms of the degree of spatial correlation and the degree of polarization between pairs of polarized photons in the scattering medium can be studied by means of the angular spectrum based on coherent wave propagation of scattered PPPs. Additionally, the suppression ability of multiply scattered photons by the spatial coherence and polarization gatings is also studied. Thus, the performance on axial resolution of the ZLSCM in the scattering medium is improved. In addition, Sheppard et al. have discussed spherical aberration and axial imaging in confocal microscopy (Sheppard et al., 1994), and conclude that the spherical aberration due to the refraction index mismatch can be successfully compensated for by altering the effective tube length of the objective and by changing the objective aperture size (Sheppard & Gu, 1991; Gu & Sheppard, 1994; Sheppard et al., 1994). Owing to the common-path feature of PPPs, which gives rise to the heterodyne signal in the ZLSCM, we anticipate that it is also able to compensate for some spherical aberrations, because the polarized pair of photons encounter similar distortions in their journey. Consequently, they offset each other during heterodyne detection. Therefore, the wavefront distortion of the scattered PPPs, both through the scattering effect of the medium and spherical aberration due to refractive index mismatch at air–glass–medium interfaces, can be theoretically reduced by way of optical heterodyne interference. Experimental results have confirmed this idea at a low concentration of the scattering medium (Chang et al., 2005). In this work, the angular distributions of the PPPs through scattering media of different concentrations are measured. In order to evaluate the performance of the imaging sectioning ability in the scattering medium, the axial resolutions at different concentrations are also measured. In the following sections, we present the experimental correlation between the angular distribution and the three gatings, the spatial filtering gating, the polarization gating and the spatial coherence gating, of the ZLSCM. The correlation between the axial resolution and the three gatings in the scattering is discussed. The correlation between the angular distribution and axial resolution is analysed at the same time.
Experimental setup and results
Figure 1 is a schematic of the ZLSCM, in which the light source is a linear polarized Zeeman He-Ne laser (HP5517A). The beam contains two orthogonal linear polarized P and S waves at a wavelength of 632.8 nm with a frequency difference of 1.6 MHz. The output power is 0.4 mW and the beam is 6 mm in diameter. The first microscope objective shown in Fig. 1 focuses the laser beam on the surface of the object (a perfect mirror), which is immersed in a turbid medium, and the backscattered light from the mirror is collected by the same objective. Thereafter, the laser beam is directed to objective 2 via a beam splitter and an analyser to collect the weakly scattered PPPs in the scattering medium. Finally, the weakly scattered photons are filtered by a pinhole and then detected by a photomultiplier tube behind the pinhole. In this setup, two identical microscope objectives (LMPLFL Olympus, 20×) with a long working distance of 12 mm and a numerical aperture of 0.4 are used.
To visualize the ability to filter out the multiple scattered photons, we first compared the angular distribution of the detected scattered PPPs from the ZLSCM with those of a CCM. The angular distribution was measured by moving the pinhole (10 µm in diameter) laterally across the back focal plane of objective 2, with polystyrene microsphere suspensions at different concentrations in triple-distilled water as the scattering medium. The diameter of the microspheres was 1.053 µm (Polysciences Inc., 07310). Figure 2 shows the measured results for the ZLSCM when the perfect mirror was immersed in the above mentioned scattering medium at 4% of volume concentration, where the optical thickness τ = µs × 2l equals 8; furthermore, µs is the scattering coefficient and l = 1000 µm, which is the thickness of the scattering medium from the inner front surface of the cuvette to the perfect mirror in this experiment. It is found that the angular distribution of ZLSCM is much narrower than that of a CCM under the configuration of a CCM when the bandpass filter and the spectrum analyser (SA, Advantest R3361A) are removed (see Fig. 1). Instead, a digital voltmeter (HP 34401 A) operated in DC mode was arranged in order to measure the light intensity. This result can be explained with the help of Fig. 3, which is essentially the same scheme as Fig. 1, but with all the essential optical elements in series for easier understanding. Owing to the multiple scattering events occurring in the scattering medium, photons starting their journey from object point A in Fig. 3 will not converge to its conjugate image point A′, but spread out over a certain area around A′. However, some weakly scattered PPPs that suffer fewer collisions in the scattering medium land on a very small area around point A′. In terms of wave propagation, these PPPs maintain an approximately spherical wavefront centring at the object point A and finally converge approximately to point A′. Therefore, the phase and polarization state of each of these PPPs remain correlated and yield a heterodyne signal. On the other hand, the multiply scattered PPPs are decorrelated on spatial coherence and degree of polarization by the multiple scattering events. As a result, even though some of them may also land at point A′, they do not contribute to the output heterodyne signal. In general, the further away from A′ the scattered PPPs land, the more scattering events they suffer. Because of this, each of these PPPs becomes more decorrelated from each other, and thus they produce a very weak heterodyne signal. Accordingly, the intensity of the detected heterodyne signal drops rapidly as we move the pinhole laterally away from the proper conjugated image point A′. In contrast, with a CCM, no heterodyne technique is adopted and thus a nonzero background signal is detected over a fixed larger area on the focal plane as we move the pinhole laterally. Thus, a ZLSCM demonstrates a much sharper angular distribution than a CCM, as shown in Fig. 2.
According to this model, as the scattering medium becomes more concentrated, PPPs get more chance to be scattered in the medium and show less probability of landing on the conjugated image point A′. Consequently, as the concentration of the scattering medium becomes higher, in addition to a decrease in the heterodyne signal, the angular distribution of the ZLSCM is expected to be narrower. Figure 4 shows the measured results for the angular distribution using microsphere suspensions of various concentrations, namely 1%, 2% and 4% by volume. It is found from these experimental results that the angular distribution of the ZLSCM stays nearly the same as the concentration of the scattering medium is increased to 4%. This constant behaviour of the angular distribution is due to the fact that most of the decorrelated scattered PPPs are suppressed by the spatial coherence gating and the polarization gating at these concentrations of the scattering medium. Therefore, the pinhole aperture dominates the angular distribution of the weakly scattered PPPs in the scattering medium. Unfortunately, the concentration in this experiment could not be increased further because the heterodyne signal began to fluctuate too much to be detected properly when the volume concentration of the scattering medium was raised higher than 4%. This was because the available output power of the Zeeman laser was too low in this measurement. Therefore, if a higher output power of the Zeeman laser or a two-frequency laser (Su et al., 1996) is available, a narrow angular distribution would be anticipated according to the previous analysis. During these measurements, the signal level of the optical heterodyne signal is low in a higher concentration of scattering medium. However, enhancement in SNR of the signal is obtainable in this setup because a common phase noise rejection mode and a narrow bandpass filter are provided. This results in coherent detection of the heterodyne signal that is contributed by weakly scattered PPPs. In this experiment, the SNR of the detected heterodyne signal at different scattering concentrations was in the range 3 < SNR < 10. This agrees with the theoretical calculation according to the analysis of
in the confocal microscope (Sheppard et al., 1995) where the quantum efficiency QE of the photomultiplier tube was 8% and the number of detected photons was np ≥ 200 per measurement. nn is defined as the electronic noise of the photomultiplier tube which nn ≈ 0 is applied accordingly.
The ability of our ZLSCM to filter out multiply scattered photons is further demonstrated by measurement of its axial resolution. The measurement was carried out by keeping the pinhole still but moving the perfect mirror longitudinally through focus by a computer-controlled microposition translation stage. The axial resolution, denoted by 2Z1/2, is defined as the full width at half maximum of the measured intensity as the perfect mirror moves back and forth through the focal plane of the first objective. To compare the performance of the ZLSCM and CCM, Fig. 5 shows the axial resolution of a CCM using a 20× objective with a 10-µm pinhole and that of a ZLSCM using the same objective with a 10-µm and a 5-µm pinhole, respectively. The concentration of the scattering medium in these experiments was 4% by volume. It can be seen in Fig. 5 that the axial resolution of the ZLSCM is much better than that of the CCM. Furthermore, Fig. 6 shows that the axial resolution of the CCM degrades monotonically as the concentration of the scattering medium becomes higher, whereas that of the ZLSCM stays practically the same for both pinhole sizes. A larger axial resolution for the CCM implies that the photomultiplier tube picks up multiply scattered photons with a larger scattered angle and then produces a detected signal because no heterodyne technique is applied. In contrast, the ZLSCM has the ability to discriminate the multiple scattered photons from the weakly scattered PPPs, which have suffered fewer collisions and therefore maintain the polarization correlation and the spatial coherence between the pair of photons. Also, the direction of propagation of weakly scattered PPPs is close to the optical axis in the scattering medium. Finally, these weakly scattered PPPs contribute the heterodyne signal. Thus, the ability of the ZLSCM to provide better axial resolution than the CCM in a scattering medium is confirmed by this experiment. It can be seen in Fig. 5 that there are oscillations in the detected intensity curve that is shown in the axial resolution of the CCM. It is believed that they are due to the spherical aberration within the system (Sheppard & Gu, 1991; Sheppard et al., 1994), which is generated by the refractive index mismatch between the air and the sample. In addition, the scattering effect of the medium also distorts the wavefront at the same time. The oscillations due to spherical aberration can be compensated for by altering the effective tube length of confocal microscope system (Sheppard et al., 1994). Figure 6 shows the results for the axial resolution of the ZLSCM both in nonscattering medium and in the scattering medium at various volume concentrations ranging over 1%, 2%, 3% and 4% of the polystyrene microspheres. It should be emphasized when the ZLSCM is used to look into a nonscattering medium, this means that the laser beam is focused on a perfect mirror that is placed in an empty cuvette of 1-cm thickness. However, it should be mentioned that, in order to prevent the back-reflected laser beam from entering into the laser cavity and inducing instability, we have always placed another identical cuvette containing 0.2% intralipid/10% AB solution (Fresenius Kabi AB, Uppsala, Sweden) just behind the Zeeman laser in order to stabilize the laser output intensity. We treat this condition as a quasi-nonscattering medium in this experiment. The relationship between the axial resolution of the ZLSCM and the size of the pinhole is shown in Fig. 6. The larger the pinhole, the worse the axial resolution becomes. The measurement with a 5-µm pinhole resulted in Z1/2= 7.5 µm, whereas a 10-µm pinhole resulted in Z1/2 = 24 µm for the quasi-nonscattering situation as defined above. We would like to emphasize that for a scattering medium, the two groups of axial resolution values for the pinhole diameters of 5 µm and 10 µm in the ZLSCM are not equal. Each group, however, stays practically the same up to 4% by volume concentration of polystyrene. However, the differences between axial resolutions with respect to the two pinhole sizes at different concentrations of the scattering medium remain the same, as shown in Fig. 6. These results imply that the pinhole size determines the axial resolution, whereas the polarization gating and the spatial coherence gating effectively suppress the multiply scattered photons. Therefore, a smaller pinhole size generally results in a better axial resolution by the ZLSCM in the scattering medium. This is similar to the CCM in a nonscattering medium.
In order to compare our results with that of the Wilson & Garlini (1987) calculation for an aberration-free CCM looking into a nonscattering medium, the calculation was carried out under paraxial approximation and we express our result in terms of the following two parameters:
where vp is the pinhole size of the optical unit, 2u1/2 is the axial resolution of the optical unit, Z1/2 denotes one half of the full width at half-maximum of the measured intensity, rp is the radius of the pinhole, sin α equals the numerical aperture of the objective in air, and λ is the wavelength of the laser. In the experiments, vp = 7.45 when a 20× objective (with an effective numerical aperture of 0.3) was combined with a 5-µm-diameter pinhole, whereas vp = 14.9 when it was combined with a 10-µm-diameter pinhole. The two measured axial resolutions of Z1/2 = 7.5 µm and 24 µm expressed by the optical units (Wilson & Garlini, 1987) are u1/2 ≈ 6.7 and 21.4, respectively. To show the performance of our ZLSCM, we compare in Fig. 7 our measured u1/2 under various scattering conditions with Wilson's paraxial approximation calculation of a CCM under aberration-free nonscattering conditions (Wilson & Garlini, 1987). It is found that the measured result of u1/2 for vp = 7.45 under quasi-nonscattering conditions falls very close to Wilson's theoretical value. Furthermore, when the scattering medium is changed to a higher-scattering condition such as the quasi-nonscattering condition plus a cuvette containing 4% by volume of polystyrene microspheres at vp = 7.45, the measured value of u1/2 is still not too far from Wilson's prediction. These results imply that the ZLSCM is not only able to reduce the scattering effect, but is also able to reduce the spherical aberration induced by the refractive index mismatch of the air–medium interface. However, the measured u1/2 values under the same scattering conditions but with a different pinhole size (10 µm diameter) of vp = 14.9 move significantly away from Wilson's calculation. Therefore, these experimental results imply that the 5-µm pinhole is more effective than the 10-µm pinhole in filtering out the weakly scattered PPPs. In other words, the pinhole size determines the effective scattering angle within which the weakly scattered PPPs are detected in the ZLSCM. Thus, when the pinhole size is decreased, the weakly scattered PPPs with larger scattered angles will not be detected. This explains why the axial resolution of our ZLSCM in Fig. 6 is quite insensitive to the concentration of the scattering medium but is very sensitive to pinhole size. According to Wilson's theory, the axial resolution 2u1/2 in the optical unit of an aberration-free nonscattering CCM with vp = 14.9 is 24; in comparison, the value of 2u1/2 with the CCM setup with quasi-nonscattering medium is found to be ≈ 160.8 in the experiment with a 10-µm pinhole (vp = 14.9) (Fig. 6). This is due to the large number of multiply scattered photons in the scattering medium that are recorded because the only gating to suppress the multiply scattered photons in this setup is the pinhole aperture. To summarize, in the ZLSCM, the overall suppressing power of both multiply scattered photons and the large scattered angle of PPPs can be treated as the product of the rejecting powers of these three individual gatings.
To examine the tomographic imaging capability of this ZLSCM, a highly reflective planar letter ‘M’, 1100 µm × 750 µm, was imaged. The letter ‘M’ was placed in the scattering medium at a distance of 1000 µm from the glass–medium interface of the cuvette, and the scattering medium was a suspension of polystyrene microspheres (0.992 µm in diameter) in triple-distilled water at a concentration of 2% by volume. The optical thickness of the medium was τ = 4. The object letter ‘M’ was created by ablating a polycarbonate surface and then coating this with a gold film to improve its uniformity and reflectivity. Measured by a surface profiler (New View5000, ZYGO), the letter ‘M’ was found to be etched 25 µm deep into the polycarbonate substrate, as shown in Fig. 8(a). In this experiment, we focused the microscope objective L1 (effective numerical aperture = 0.3) onto the unablated surface of the substrate and scanned it by moving the sample laterally. A 10-µm pinhole was used for this measurement. Figure 8(b) shows the tomographic image of the letter ‘M’. The image consists of 110 × 75 scanned dots. In the scanning, the distance of each scanning step was 10 µm. The image in Fig. 8(b) shows clearly that the areas other than the letter ‘M’ are in focus and the intensity drops to half of its maximum value in the area occupied by the letter ‘M’. This result is expected, because this area is out of focus by 25 µm and the axial resolution of this ZLSCM is approximately 58 µm. Thus, if the letter ‘M’ were scanned with a 5-µm pinhole, the scanned image would be even better, because the axial resolution becomes better than with a 10-µm pinhole. As seen in Fig. 8, the intensity distribution of the letter ‘M’ is not uniform; this is caused by the surface roughness in the area of the letter.
We have presented a ZLSCM that employs a two-frequency Zeeman laser as the light source. The microscope is designed to produce a heterodyne signal in a scanning manner. Accordingly, the microscope uses three gatings simultaneously: the spatial coherence gating, the polarization gating, and the spatial filtering gating. The function of the first two gatings is to discriminate the multiply scattered photons from the weakly scattered PPPs that produce the heterodyne signal and carry the object information to the image plane. The function of the third gating is to further exclude large scattering angle PPPs, which would land closely around the perfect image point if a smaller pinhole size was introduced into the ZLSCM. Although these PPPs have been distorted due to collisions with the scattering medium, the distortion is not so serious. They are still able to give rise to a heterodyne signal. This fact is confirmed by the experimental results in Figs 4 and 6, where a smaller pinhole gives a better angular distribution and axial resolution simultaneously. Because the measuring beam in this microscope is derived from a Zeeman laser, the two polarization components of the measuring beam travel along a common path in the scattering medium. In addition to the functions of the three gatings, the common-path feature results in the ZLSCM having a particulr feature: it enables the scattered PPPs to contribute to the heterodyne signal coherently. Thus, it gives rise to a higher SNR and a higher modulation index based on the detected heterodyne signal, which results in better sensitivity of detection. As a result, the ZLSCM can be used to look deeper into a scattering medium, or alternatively to look through a denser scattering medium. As shown in Fig. 6, although the axial resolution of the microscope is degraded somewhat when there is a change from a quasi-nonscattering medium to a scattering medium, as expected, it maintains a nearly constant value as the concentration of the scattering medium is increased to a value up to 4% by volume of polystyrene microspheres. This result would not be obtainable if the multiply scattered photons were not suppressed effectively. Furthermore, when dealing with a quasi-nonscattering medium as described in the text, and when using a 5-µm pinhole in our system, the axial resolution of the ZLSCM is very close to the theoretical value of Wilson's theory for an aberration-free and nonscattering situation. This result implies that the ZLSCM is able to reduce the spherical aberration caused by the refractive index mismatch or induced by scattering of the turbid medium in which the examined object is immersed. This is because of the common-path feature, whereby the P-polarized waves and the S-polarized waves suffer the same distortion in the wavefront. Consequently, these distortions in the P waves and the S waves offset each other, and therefore the wavefront distortion due to these effects is reduced by heterodyne interference. Other than that, the common phase rejection mode of PPPs can also reduce the background phase noise, including laser frequency noise, and the optical path-dependent phase noise at the same time. In addition, the offset of optical path dependent phase of scattered PPP results in coherent detection of the heterodyne signal. Thus, the SNR of detected signal is further enhanced. Therefore, the performance of the ZLSCM in the scattering medium is similar to that of the CCM in nonscattering medium qualitatively. The pinhole aperture plays a dominant role to the performance on axial resolution of ZLSCM in the scattering medium. This is different from the situation with a conventional optical heterodyne microscope in the scattering medium, in that the reference beam strengthens the SNR of the heterodyne signal via a different optical path in the interferometer (Kempe et al., 1997).
Fundamentally, the mechanism of the ZLSCM is apparently different from that of the LCM and CCM. The ZLSCM enables the recovery of the tomographic image in the scattering medium via the previously mentioned three gatings simultaneously, even though the signal level is much lower than with the conventional confocal reflectance microscope (Gu et al., 1991; Kimura & Wilson, 1991; Sung & Richards-Kortum, 2003) and the confocal fluorescence microscope (Hell et al., 1992). However, the performance of the ZLSCM based on reflectance microscopy with heterodyne interference in conjunction with the spatial coherence, polarization and spatial filtering gatings is similar to that of the optical coherence microscope (Izatt et al., 1994), which is able to recover the tomographic image in the scattering medium properly. When a higher output power of the two-frequency laser is available, this proposed two-frequency laser scanning confocal microscope is anticipated to be able to demonstrate sectioning imaging capability in a higher scattering medium. Potentially, the performance of the ZLSCM could become comparable to that of the conventional confocal reflectance microscope in nonscattering medium.
This research was partially supported by the National Science Council of Taiwan through grant no. NSC 92-2215-E-010-001.