1. Introduction

Visualization of cellular and sub-cellular structures and processes in intact tissues is highly desirable in biological research [1–3]. Light microscopy, with submicron spatial resolution, has been an indispensible and powerful tool in various fields of biological research and medical diagnoses. Biological tissues are heterogeneous from microscopic to macroscopic scales. Generally they appear opaque to visible and near infrared light as photons are subject to strong scattering and absorption. Scattering, especially multiple scattering, is an undesirable phenomenon in imaging science that alters the propagation direction of photons. The scattering mean free path l_s, the average distance between consecutive scattering events, has a typical value around 100 microns in human soft tissues [4]. While conventional wide field microscopy can only deal with very thin samples, the invention of confocal microscopy (CM) was a significant advance in modern light microscopy as optical sectioning capability is provided [5–8]. In a confocal microscope, the sample is illuminated by a focused beam and scanned point-by-point. The signal from the focal point is descanned and collected by a photodetector behind a confocal pinhole. In an ideal situation, out-of-focus light from the sample is mostly rejected. However, this selective detection scheme is not so effective when the focal point moves into the sample for such a depth that the scattered photons dominate the ballistic ones. The point spread function, which determines the spatial resolution, is broadened rapidly in space with increasing imaging depth [9]. Although a strong target might be detectable at depths over a few l_s, high resolution details can be easily masked by background signal even when the target is located at only one l_s from the surface. Subcellular imaging with a confocal microscope is usually performed at an imaging depth up to a few tens of microns. To maintain a near-diffraction-limited resolution in a deeper region, it is essential to develop a mechanism to effectively prevent the multiply scattered photons from being involved in image formation.

In multi-photon microscopy (MPM), the focused illumination beam is further concentrated within an ultra short time window of less than one picosecond [10–12]. The nonlinear absorption rate decays rapidly out of focus and this selective excitation method is effective when the imaging depth is less than 1 mm [9]. Recently differential aberration two-photon microscopy was developed to further reject the background. However, no significantly improved imaging depth has been reported [13]. MPM has become an increasingly popular alternative to confocal microscopy as a result of improved imaging depth and localized photochemistry. However, MPM is a very expensive technique that uses laser sources of ultra short pulses. In addition, single photon excitation is preferred over multi-photon excitation in some situations in which nonlinear photo-damage and availability of fluorescence probes are of concern.

The coherence gating mechanism in optical coherence tomography (OCT) is very effective in picking up the desired signal [14, 15]. Although some of multiply scattered light still reaches the photodetector, only the back scattered (scattered once) or reflected light, which has a well defined optical pathlength and polarization state, generates the fringe signal for image formation. The imaging depth and speed have been further improved recently with the Fourier domain techniques [16, 17]. A cross-sectional scanning rate over 30 frames per second is readily achievable with an imaging depth up to 3 mm. Unfortunately OCT is not compatible with fluorescence. While it has been successfully applied to in vivo visualization of the fine structures of human eyes and hollow organs, its molecular imaging capability is rather limited [18–21].

Photoacoustic tomography is a fast developing technique for molecular imaging of deep tissue beyond a few millimeters [22]. The spatial resolution is determined by the wavelength of optically stimulated ultrasound waves and is generally limited to tens of microns even with a high frequency transducer. Fluorescence based molecular tomography [23] or diffuse optical tomography [24] is capable of measuring spatial distributions of chromophores within large human organs or small animal models. Of course, the associated spatial resolution is usually limited to a few millimeters.

We therefore developed focal modulation microscopy, a novel technique that targets an imaging depth greater than 0.5 mm combined with diffraction limited spatial resolution and molecular specificity. By the use of a spatial phase modulator in the excitation light path, an intensity modulation is achieved mainly in the focal volume only, even when the focal point is located deep inside a turbid medium. The oscillatory component in the detected fluorescence signal can be readily differentiated from background signal caused by multiple scattering. Our implementation allows simultaneous acquisition of confocal and FMM images. We have demonstrated the advantages of FMM over CM with a series of image experiments using cartilage tissue from chicken. An imaging depth of 600 microns has been achieved with FMM.

2. Method

Shown in Fig. 1 is the schematic diagram of a prototype FMM system developed at the Optical Bioimaging Lab, National University of Singapore. It is very similar to a confocal microscope. However, the key difference is that a spatial phase modulator is inserted into the excitation light path. The light source is a 640 nm solid state single frequency laser (RCL-025-640-S, CrystaLaser), whose 25 mW output beam is expanded to about 5 mm in diameter. When passing through the spatial phase modulator, the beam is split into two spatially separated half-beams, which are parallel and subject to different phase delays. Currently spatial phase modulation is implemented with two parallel mirrors (M1 and M2) inside the dashed box. M1 is mounted on a stationary base while M2 is mounted on a piezoelectric actuator (AE0203D04, Thorlabs Inc.). A sinusoidal voltage signal of a single frequency f=5 kHz superimposed on an appropriate dc bias is applied to the actuator to vary the relative phase shift periodically between 0 and π. The spatial phase modulated excitation beam is mostly (~95%) reflected by a dichroic mirror (DM, XF2035, Omega Optical Inc.) and is then directed by a two-dimensional fast steering mirror (FSM-300-01, Newport Inc.) to a 20X objective (LUCPLFLN 20X, Olympus Inc). Due to a varying spatial phase distribution, the excitation beam entering the objective aperture does not necessarily converge to the focal point. Consequently, an intensity modulation of the excitation light is achieved around the focal point. When the focal point is within a turbid medium, the excitation photons reaching the focal plane include both ballistic and scattered photons. Only the ballistic photons contribute to oscillatory excitation rates as they have well defined phase and polarization. The oscillatory excitation rate is in-phase with the modulation signal inside a small focal volume and out-of-phase elsewhere. Fluorescence emission is collected by the same objective and descanned is performed by the same fast steering mirror. A long pass filter (3RD670LP, Omega Optical Inc.) is used to further reject the excitation light already suppressed by the DM. The cut-on wavelength for both the DM and the long pass filter is between 667 and 670 nm. Then the fluorescence light is focused with an achromat and coupled into a single mode (SM) optic fiber (P3-630A-FC-5, Thorlabs Inc.), which functions as a detection pinhole. A photomultiplier tube (R928, Hamamatsu Photonics Co.) converts the weak light signal to an electrical signal, which is further enhanced by a 40 dB amplifier before being digitized into a personal computer (PC). The acquired photoelectrical signal contains a dc component, an ac component at 5 kHz due to modulated excitation, and random noise. A Fast Fourier Transform (FFT) is performed on the PC to retrieve both ac and dc signals. The sum of the ac amplitude and the dc magnitude is equal to the maximum emission intensity and thus equivalent to the conventional CM signal. FMM, however, uses the ac amplitude only for image formation. The PC also controls the 2D fast steering mirror to scan the sample point-to-point to obtain both CM and FMM images simultaneously.

 

Fig. 1. Schematic diagram of the prototype focal modulation microscopy system. The spatial phase distribution of the 640 nm excitation beam (red) is modulated by the use of two parallel mirrors, M1 (stationary) and M2 (oscillating axially at 5 kHz). The fluorescence emission (purple) from the focal volume is collected by a fiber based confocal detection system, and then the oscillatory component at 5 kHz is retrieved for image formation. The personal computer is used for data acquisition and analysis, lateral scanning with the fast steering mirror, and axial scanning with the 3D stage.

Download Full Size | PDF

 

Fig. 2. (a) Two-dimensional distribution of the differential excitation power in the focal plane. (b) Confocal signal (blue), FMM signal (green), and modulation depth (red) as functions of the normalized pinhole size.

Download Full Size | PDF

Spatial phase modulation has been used in nonlinear microscopy for background rejection [13] and axial resolution improvement [25]. However, one photon fluorescence microscopy is intrinsically different from nonlinear microscopy due to different dependence of excitation rate on light intensity. Shown in Fig. 2 (a) is the two-dimensional distribution of the differential excitation power (obtained by simulation) normalized to its peak value. The oscillatory fluorescence emissions from different regions are either in-phase (with positive values) or out-of-phase (with negative values) with the modulation signal. If all the emissions from the focal plane are collected, the ac component will vanish. This is a feature very different from nonlinear excitation. To maximize the FMM signal, a selective detection scheme is necessary to pick up the in-phase signal only. The confocal pinhole defines a small focal volume for detection from which emissions are collected. Out-of-phase oscillatory emissions, mainly originating from regions out of the focal volume (the dark blue region in Fig. 2 (a)), are suppressed and kept from adding to the FMM signal destructively when a pinhole of appropriate size is used. Fig. 2 (b) shows numerical simulation results on the relationships of FMM and confocal signals with the pinhole size. The pinhole radius a is normalized as ν=2π Na/λ, where λ is emission wavelength and N is the numerical aperture of the achromat. As expected, the confocal signal increases monotonically with increasing pinhole size (blue curve) while the FMM signal (peak to peak) decreases monotonically after reaching its maximum around v=3 (green curve). The modulation depth, defined as the ratio of the peak to peak FMM signal to the confocal signal, always becomes smaller with a larger pinhole (red curve). The SM fiber we are using has a mode diameter around 4.3 microns (ν≈1.9) and the measured modulation depth is roughly 70%, agreeing with the numerical simulation result. The modulation depth will be reduced to less than 50% if another SM fiber (9 microns in mode diameter) is used and becomes much smaller with a multimode fiber. Similar to its function in confocal microscopy, the pinhole also provides substantial rejection of out-of focus background in the first place so as to achieve a much lower shot noise floor than those of wide-field microscopy methods.

3. Experiments and results

To demonstrate the capability of our method for in vivo imaging of cellular and sub-cellular structure and function, we used chicken cartilage as a sample tissue to evaluate the performance of FMM. Chondrocytes are the only cells found in cartilage. The cells are usually of a rounded or bluntly angular form, lying in groups of two or more in a glandular or almost homogeneous matrix. Chicken cartilage was cut into slices around 1 mm in thickness and stained with DiD (DiIC18(5), Invitrogen Corp.), a lipophilic tracer for cell membrane labeling. The laser power was attenuated by 10-1000 times to avoid fast photobleaching. Figures 3 (a) and 3 (b) are CM and FMM images acquired at a depth around 280 microns. One can easily see that optical sectioning is not effective in the confocal image (Fig. 3 (a)). Fluorescence signals from layers other than the focal plane cast shadows in the image, resulting in overlapping structures and blurred cellular shapes. On the contrary, the FMM image (Fig. 3 (b)) shows uncompromised quality. The upper-central area was then scanned with a four times finer step (about 100 nm) and the high magnification images are displayed in Figs. 3 (c) and 3 (d). The FMM image (Fig. 3 (d)) provides detailed information with sub-micron spatial resolution and excellent contrast, which are not available from the CM image (Fig. 3 (c)). We then pushed further to test the maximal penetration depth of our prototype FMM system. Shown in Fig. 3 (e) and 3 (f) are FMM images from 500 and 600 microns in depth. It is evident that resolution is still high enough to visualize cellular structures. Beyond 600 microns the shot noise associated with the background started to overwhelm the FMM signal.

To the best of our knowledge, the penetration depth we report in this paper has never been achieved by one photon fluorescence based microscopy when imaging optically dense biological tissues. However, there might be still room for further improvements. The tissue samples in our experiments were uniformly stained. As a result, the superficial layers strongly absorbed the excitation light and generated intense background emissions. For certain applications where only the deep region of interest is stained, an even larger penetration depth should be possible. More photostable and brighter fluorescent labels (e.g., quantum dots) can also help further increase the imaging depth.

 

Fig. 3. Fluorescence images of chondrocytes obtained from chicken cartilage. Confocal images ((a) and (c)) were acquired simultaneously with the corresponding FMM images ((b) and (d)) at a depth of 280 microns. (e) and (f) are FMM images obtained from 500 and 600 microns in depth.

Download Full Size | PDF

4. Discussions and conclusion

The imaging depth demonstrated in this paper is related to a relatively long excitation wavelength of 640 nm. Generally the scattering coefficient of biological tissue decreases with increasing wavelength, and in many situations the attenuation of ballistic light is dominated by scattering. For excitation light of shorter wavelength, such as UV or blue light, the achievable imaging depth should be smaller. On the other hand, near infrared fluorescence dyes (Indocyanine Green, for example) would allow the use of an even longer excitation wavelength (780 nm) and deeper imaging. In our future research, we will systematically study the dependence of imaging depth on excitation wavelength and fluorescence dye properties. The imaging speed of our prototype FMM is limited by the modulation frequency. The minimal dwelling time at each pixel is 0.2 ms, significantly longer than the typical 10 microsecond pixel dwelling time for commercial CM and MPM systems. We plan to develop electro-optical material based spatial phase modulators that can be modulated at a frequency more than a few MHz. Near-real-time FMM imaging (a few frames per second) would be possible with such a modulator.

To conclude, we have developed and experimentally demonstrated a novel microscopy method for molecular imaging of thick biological tissues with one photon excited fluorescence. We believe that FMM will find numerous applications in basic biological research and clinical diagnoses with further improvement in imaging speed and compatibility with more excitation wavelengths and fluorescence dyes.