2.1.

Experimental Setup

The general schematic of the iTS microscope setup is depicted in Fig. 1. In this paper, two different types of pulsed laser sources were chosen for different iTS microscopy demonstrations: (1) source A: a home-built fiber mode-locked laser (center $\text{wavelength}=1064\mathrm{nm}$, $\text{bandwidth}=10\mathrm{nm}$, $\text{repetition rate}=26.3\mathrm{MHz}$) with preamplification by a home-made ytterbium-doped fiber amplifier (YDFA) ($\text{net optical gain}=\sim 6\mathrm{dB}$), which is then prechirped by a 5-km single-mode fiber (SMF, Nufern 1060XP) and further amplified by a YDFA (with net optical gain of $\sim 25\mathrm{dB}$)¹⁸ and (2) source B: a broadband pulsed laser ($\text{bandwidth}=60\mathrm{nm}$) generated through supercontinuum generation in a highly nonlinear fiber (431C, $\text{length}=\sim 100\mathrm{m}$), pumped by a pulsed laser (Time-bandwidth Products Ltd., Switzerland, center $\text{wavelength}=1064\mathrm{nm}$, $\text{repetition rate}=20\mathrm{MHz}$). The pulsed laser beam is first split into two paths through a beam splitter ($\text{power splitting ratio}=55:45$), forming a reference path and a sample path in a Michelson interferometer configuration. A one-dimensional (1-D) spectral shower is diffracted by passing the sample beam through a diffraction grating (Richardson Gratings Ltd., Rochester, New York, $\text{groove density}=1200\text{lines}/\mathrm{mm}$, $\mathrm{Littrow}\text{angle}=42.1\mathrm{deg}$) and illuminates the samples. By a double-pass transmission configuration using two objective lenses (both with $\mathrm{NA}=0.66$), the spatial coordinates are mapped to the wavelength spectrum of the 1-D spectral shower. This is a process called spectral encoding.^19,20 The spectrally encoded pulsed beam from the sample arm is then interfered with its uncoded replica from the reference arm by the same beam splitter. The combined beam is then coupled into a dispersive fiber by a fiber collimator. The interfered pulses undergo optical time-stretch (also known as dispersive Fourier transform^21,22) along the dispersive fiber through group velocity dispersion (GVD). The dispersive fiber modules used with sources A and B are different: (1) a 13-km few-mode fiber (FMF, SMF28) with a total GVD value of $\sim 0.45\mathrm{ns}/\mathrm{nm}$ for source A and (2) a 5-km SMF (Nufern 1060XP) and a 6-km FMF (SMF28) with total GVD value of $\sim 0.35\mathrm{ns}/\mathrm{nm}$ for source B. FMF has been found to be an effective dispersive fiber option for the time-stretch process with a selective mode excitation.²³ With these GVDs, the time-stretched interferograms can be completely resolved and the diffraction limited resolution is preserved.^12,17 The time-stretch interferograms are also amplified by an in-line YDFA, which achieves an on-off gain as high as $\sim 15\mathrm{dB}$ in order to compensate for the dispersive loss and, thus, enhance the imaging sensitivity. Afterward, a high-speed ($\text{electrical bandwidth}=8\mathrm{GHz}$) single-pixel photo-detector with a sensitivity ($\mathrm{NEP}=-19\mathrm{dBm}$) and a real-time oscilloscope ($\text{electrical bandwidth}=16.8\mathrm{GHz}$, $\text{sampling rate}=80\mathrm{GS}/\mathrm{s}$) is used to digitize and acquire the temporal interferograms. To acquire a two-dimensional (2-D) image, the sample is line-scanned in the direction orthogonal to the spectral shower’s main axis in the case of imaging fixed cells and tissues, or the unidirectional flow of the cells naturally provides the orthogonal image scan in the case of microfluidic flow imaging.

Fig. 1

Interferometric time-stretch (iTS) microscope setup. (a) A broadband pulsed laser is used for generating a one-dimensional spectral shower after passing through a diffraction grating, which is then focused onto the sample plane by an objective lens. Built upon a Michelson interferometer configuration, the setup recombines the double-passed, spectrally encoded sample beam (S) with the reference beam (R) by the beam splitter. The interfered pulsed beams are coupled into the dispersive fiber by the fiber collimator. Time-stretch takes place in the dispersive fiber. An in-line amplifier is added in order to compensate the dispersive loss. Note that the group velocity dispersion should be large enough to provide sufficient time-stretch for preserving the diffraction-limited resolution. The time-stretched interferograms are detected by the single-pixel photodetector and a real-time oscilloscope. Phase retrieval and unwrapping are done off-line to reconstruct the quantitative phase information. (b) The corresponding temporal and spectral waveforms in different stages [steps 1 to 4 as shown in (a)]. R and S refer to the reference and the sample arms in a Michelson interferometer configuration.

2.2.

Image Reconstruction of iTS Microscopy

The quantitative phase image of the sample is retrieved in a way similar to conventional QPI.²⁴ In the case of reconstructing the iTS images, an additional step is required, i.e., transformation of the temporal interferograms into the spatial ones. Such mapping between space and time can be calibrated through the relation²⁵ $\delta x=\delta t\times C/\mathrm{GVD}$, where $\delta x$ and $\delta t$ correspond to the changes in space and time, respectively. $C$ is the conversion factor between the space and wavelength that is governed by the spatial dispersion law of the diffraction grating.²⁵ Once the mapping is known, the temporal interferograms obtained from the oscilloscope can be expressed in the following spatial form:²⁶

Since the phase term retrieved from the tangent operation has limits between $-\pi$ and $+\pi$, a 2-D phase unwrapping algorithm, Goldstein’s algorithm, is applied to obtain the unwrapped true phase from the discontinuous wrapped phase. Such 2-D phase unwrapping of a $200\times 3000$ points image can be done within 2 min by a custom-written MATLAB® code run on a personal computer.

It should be noted that the image resolution of iTS is essentially defined in the same way as the original time-stretch microscopy, as discussed in detail elsewhere.^10,25 In brief, the resolution of iTS is governed by three different limiting regimes, each of which is dominated by (1) the spectral resolution of the diffraction grating (spatial-dispersion limited), (2) the spectral resolution defined by stationary-phase approximation (SPA) in the time-stretch process (SPA-limited), or (3) the temporal resolution of the digitizer (digitizer-limited). Each of these limiting cases has an associated spatial resolution, and the final image resolution is defined as the maximum of these three values. It is desirable that the resolution should be ultimately limited by the spatial dispersion (i.e., spectral encoding process). Based on our current system configuration and following the analysis presented in Ref. 25, our current system design yields a resolution limited by spatial dispersion and is calculated to be $\sim 1.2\mu \mathrm{m}$—sufficient for cellular imaging, as will be shown in a later section.

3.1.

Basic Performance of iTS Microscopy

To first demonstrate the quantitative imaging capability of iTS, we image an ISO 12233 resolution chart, which is fabricated by depositing a thin layer of photo-resist (MicroChem Corp., Newton, Massachusetts, Shipley 1805, refractive index $\sim 1.52$ at 1064 nm, thickness of $\sim 400\mathrm{nm}$) on a glass slide, by the iTS microscope based on source B. By taking an average of 20 single-shot line-scans, we can achieve an effective line-scan rate as high as 1 MHz. The 2-D interferogram is obtained by scanning the resolution chart in the orthogonal direction with a step size of 0.5 μm. Figures 2(a) and 2(b) show the bright-field image taken by the conventional optical microscope and the iTS microscopic image of the same area of the resolution chart. iTS can not only show a similar image quality at the ultrafast line-scan rate of 1 MHz, but can also reveal the quantitative height profile. We validate the performance of such ultrafast noncontact profilometry by comparing the height profile measured by iTS with that measured by a profilometer (Dektak XT, Bruker Corp., Germany), as shown in Fig. 2(c).

Fig. 2

Basic performance of iTS microscopy. (a) ISO 12233 resolution chart captured by bright-field light microscope. (b) Same region of resolution chart captured by iTS microscope at 1 MHz line-scan rate. The color bar shows the corresponding height. (c) Height profile along dashed line in (a) to (b) measured by profilometer (top) and iTS microscopy (bottom). (d) Tomographic image of an oil-in-water emulsion obtained by iTS microscopy at 1 MHz line-scan rate. The height profile is evaluated by the known refractive indices of water (1.33) and silicone oil ($\sim 1.40$). Inset shows the corresponding three-dimensional tomography of the emulsion droplet. (e) 1000 sampled line-scans of quantitative phase captured by the iTS microscope within 95 s. The bottom inset shows the average temporal phase stability along the spectral shower direction. The overall phase stability along the spectral shower direction is 0.01 rad, which is comparable to conventional quantitative phase imaging. Arrows indicate scanning direction.

We also image the oil-in-water emulsion droplets by iTS as another proof-of-principle experiment by using source A [Fig. 2(d)]. The droplets are generated through the flow focusing scheme in a home-made microfluidic channel and are then transferred to a glass slide covered by the coverslip. We note that high-speed quantitative characterization of droplet emulsion formation is crucial for high-throughput monitoring and screening during drug encapsulation.^27,28 With the known refractive indices of water (1.327) and silicone oil (1.40), the iTS microscope, running at a 1 MHz line-scan rate, is able to retrieve the quantitative morphology of the emulsion. The flat-top profile of the emulsion (with a thickness of $\sim 2\mu \mathrm{m}$) is attributed to the droplet being compressed by the coverslip.

To verify the stability and accuracy of the phase acquired by the current iTS microscope, we measure the phase of the 1-D iTS line-scan for a total time of 95 s at a line-scan rate of 26 MHz. The phase stability/variation is defined as the standard deviation of phase within a diffraction-limited focused spot and is found to be 0.01 rad [Fig. 2(e)]. This is comparable to the phase stability of the conventional QPI system, but is achieved at an orders-of-magnitude faster imaging speed.²⁴ Further improvement in phase stability can be obtained with a pulse laser source with better temporal power stability.

3.2.

Quantitative Cellular and Tissue Imaging by iTS Microscopy

We also demonstrate iTS imaging (using source B) of both the cell culture samples and tissue section samples mounted on the glass slides. By scanning the unstained nasopharyngeal epithelial cells in the direction orthogonal to the spectral shower with a step size of 0.5 μm, a 2-D iTS image can be acquired. The morphology of the epithelial cells [Fig. 3(a)] can be revealed by iTS microscopy when compared with the classical phase contrast images on the same region of interest [Fig. 3(b)]. More importantly, iTS imaging can also retrieve the quantitative phase profile of the cells at an ultrahigh line-scan rate of 1 MHz. We note that the extracted phase information can be used for further quantitative analysis of the characteristics of the individual cells, e.g., sphericity, cell dry mass, etc.³ Combined with a continuously running automated scanner similar to the ones adopted in WSI scanners but operated at a significantly higher speed, e.g., $>1\mathrm{m}/\mathrm{s}$, an iTS microscope could be an attractive tool for boosting the throughput of WSI of biological specimens, opening up new opportunities for big-data analysis in biomedical diagnostics as well as basic life science studies. As an example, the scanning speed of the commercially available WSI scanners is severely impeded by the speed limitation of the current camera technology. Compromising the camera speed, today’s scanners require $>1\min$ to scan a typical area $1.5\times 1.5\mathrm{cm}$ at $20\times$ magnification.² Furthermore, in some applications such as digital pathology, that rescanning is often required. As a result, a WSI system with ultrafast iTS microscopy could be of great value for realizing truly high-throughput quantitative screening.

Fig. 3

iTS microscopy of fixed cells and tissues on slides. (a) Nasopharyngeal epithelial cells captured by iTS microscopy at 1 MHz line-scan rate. (b) Same region of cells captured by conventional phase contrast microscopy. The color bar represents the quantitative phase values in radian. (c) Spleen tissue image captured by iTS microscopy at 1 MHz line-scan rate by stitching 10 iTS subimages. It is in good agreement morphologically with the bright-field image of the tissue with H&E stain. (e) Zoom-in views of the highlighted areas shown in (c) and (d). Color bars in (d) and (e) show the refractive index values. Arrows indicate scanning direction.

We further demonstrate this potential by performing large-area iTS imaging of a tissue section on a slide (using source B). By scanning the tissue in the orthogonal direction with a step size of 1 μm at a 1 MHz line-scan rate, a 2-D iTS image of the tissue section is acquired. A large-area iTS image of the tissue section ($\sim 500\mu \mathrm{m}\times 1000\mu \mathrm{m}$ is then formed by stitching 10 iTS subimages. The refractive index map of the spleen tissue quantified by iTS microscopy is shown in Fig. 3(c), which is morphologically consistent with the H&E-stained tissue captured by a bright-field optical microscope in the same area [Fig. 3(d)]. The heterogeneous refractive index distribution of spleen tissue is evaluated by taking the known refractive index of the surrounding phosphate bovine solution, which is 1.331 at 1 μm, and the known thickness of the sectioned tissue, which is 5 μm. Once again, the iTS microscope is able to reveal quantitative intrinsic information (refractive index in this case) of the tissue with cellular resolution and at a 1 MHz line-scan rate. Similar to the conventional QPI modalities, the quantitative phase image acquired by iTS microscopy can obtain not only the spatial distribution of the tissue refractive index, but also the scattering mean free path as well as the anisotropy factor.⁷ As these parameters can be regarded as the effective intrinsic biomarkers for new types of label-free disease diagnosis,⁸ iTS imaging has the unique advantage of making high-throughput WSI-based quantitative tissue histopathology possible.

3.3.

iTS Microscopy for Quantitative Cellular Imaging in Ultrafast Microfluidic Flow

Finally, we demonstrate iTS imaging of ultrafast flowing cells in the microfluidic platform. Two types of unstained and immortalized cell lines are used: human cervical cancer cells (HeLa) [Figs. 4(a) to 4(e)] and normal hepatocyte cells (MIHA) [Figs. 4(f) to 4(j)], flowing at the ultrahigh speed of 8 and $0.4\mathrm{m}/\mathrm{s}$ in the microfluidic channel, respectively. Such flow speed range corresponds to an imaging throughput as high as $\sim \mathrm{80,000}\text{cells}/\mathrm{s}$. The ultrafast fluidic flow is accomplished by the inertial flow focusing scheme, which helps align the flowing cells at the center region of the microfluidic channel (see Appendix for detailed fabrication procedures).²⁹ Enabling the ultrafast line-scan rate of 26 MHz (determined by the repetition rate of source A) as well as the ultrashort exposure time of 30 ps (determined by the time-bandwidth product of each spectrally resolvable subpulse of the spectral shower^9,10), respectively, the QPI of the fast flowing cells can be captured without image blur by iTS microscopy. This is clearly evident from the fact that the individual cells in an aggregate in such high-speed flow can be identified [see Figs. 4(a) to 4(c) and 4(f) to 4(h)]. More importantly, this is the first time, to the best of our knowledge, that heterogeneity of the subcellular morphology can also be resolved and quantified by iTS microscopy. In this ultrahigh-speed flow imaging scenario, the average phase fluctuation across 2620 line-scans was measured to be 0.00271 rad, which is practical for quantitative cellular analysis as demonstrated by the conventional QPI modalities. Operating in the cellular flow imaging scenario, iTS microscopy is a unique tool for bringing label-free quantitative phase imaging into the flow cytometry platform with an imaging throughput not achievable by any existing QPI techniques. As a result, iTS microscropy, combined with the state-of-the-art flow cytometer, could be of great value for quantitative single-cell analysis with high throughput as well as high content.

Fig. 4

Imaging of mammalian cells by iTS microscopy in ultrafast flow. (a) to (e) iTS microscopic images of stain-free HeLa cells flowing at a speed of $8\mathrm{m}/\mathrm{s}$. (a) to (c) Clusters of HeLa cells can be easily distinguished under iTS microscopy with subcellular resolution ($\sim 1.2\mu \mathrm{m}$. (d) and (e) Single HeLa cells flowing at ultrahigh speed. (f) to (j) iTS microscopic images of stain-free MIHA cells flowing at a speed of $0.4\mathrm{m}/\mathrm{s}$. (f) and (g) Clusters of MIHA cells can be easily distinguished under iTS microscopy with subcellular resolution. (h) to (j) Single MIHA cells flowing at $0.4\mathrm{m}/\mathrm{s}$. All color bars represent the phase in radian. Arrows indicate flow direction.